Chapter 1 Exercises on Descriptive Statistics and Data Types

Chapter 1
Exercise 1.1 – 1.9
1.1 Descriptive statistics is made up of those methods whose purpose is to organize and summarize a
data set. Inferential statistics refers to those procedures or techniques whose purpose is to
generalize or make an inference about the population based on the information in the sample.
1.2 A population is the entire collection of objects or individuals about which information is desired. A
sample is a subset of the population selected for study in some prescribed manner.
1.3 They are from a sample. Only some travelers were polled and the results led to estimates for the
population of interest, not the exact percentage.
1.4 The sample is the 2121 children that were in the study; the population of interest is all children aged
between the ages of one and four.
1.5 The population of interest is the entire student body (the 15,000 students). The sample consists of the
200 students interviewed.
1.6 The population consists of all 7000 property owners. The sample consists of the 500 property owners
surveyed.
1.7 The population consists of all single-family homes in Northridge. The sample consists of the 100
homes selected for inspection.
1.8 The population consists of all 2006 Mazda 6s. The sample consists of the six Mazdas of this type
selected for testing.
1.9 The population consists of all 5000 bricks in the lot. The sample consists of the 100 bricks selected
for inspection.
1.10 a. numerical (discrete)

b. categorical
c. numerical (continuous)
d. numerical (continuous)
e. categorical
1.11 a. categorical
b. categorical
c. numerical (discrete)
d. numerical (continuous)
e. categorical (each zip code identifies a geographical region)
f. numerical (continuous)
1.12 a. discrete
b. continuous
c. discrete
d. discrete
1.13 a. continuous
b. continuous
c. continuous
d. discrete
1
1.14 For example:
a. General Motors, Toyota, Aston Martin, Ford, Jaguar, …
b. 3.23, 2.92, 4.0, 2.8, …
c. 2, 0, 1, 4, 3, …
d. 49.2, 48.84, 50.3, 50.23, …
e. 10, 15.5, 17, 3, 6.5, …
1.15 a. Gender, Brand of Motorcycle and Telephone area code. b. Number of previous motorcycles
owned. c. Bar chart. d. Dotplot.
1.16
Most summer movies have box office sales of between $50 million and $152 million. There is a small
cluster of 3 films that have sales of about $200 million. The two top box office totals for the summer of
2002 were significantly higher: Star Wars, Episode II at $300.1 million and Spider-Man at
403.7million.
1.17
Breakdown of all Complaints, by Type
30%
25%
20%
15%
10%
5%
0%
Credit Card Phone or Bank Fraud Employment Other
Fraud Utilities Fraud Fraud
The most common single type of identity theft is credit card fraud with 28% of the total complaints,
followed by phone or utilities and bank fraud, both just under 20%. Employment fraud is less at 13%.
1.18 a. Categorical. b. No, because a dotplot is used for numerical data.

c.
Literacy Level in Adults as Classified by the National

Assessment of Adult Literacy
Percent of Adults
50%
40%
30%
20%
10%
0%
Below Basic Basic Intermediate Proficient
2
1.19
"Plagiarism is a significant Problem in Academic

Institutions"
160
140
Frequency 120
100
80
60
40
20
0
Strongly Disagree Not sure Agree Strongly
Disagree Agree
Response
1.20 a.
Causes of Children's Weight Problems
40%
35%
30%
25%
20%
15%
10%
5%
0%
Lack of Easy Genetics Eating Medical Overeating
Exercise access to unhealthy condition
junk food food
b. The categories “Easy access to junk food”, “Eating unhealthy food” and “Overeating” could all be
combined into a single category. They can all be described as “Poor eating habits.”
3
1.21
For both sports, there are few universities where the overall graduation rate is lower than that of the
scholarship athletes. At the schools where athletes do better than the overall rate, there are more
schools where the basketball players do well, however, there are some schools where the basketball
players graduate at a much poorer rate than overall. Overall, the athletes of both sports, graduate
about 10% lower than overall, but the range for football players is lower (35% lower to 8% higher)
than for basketball players (39% lower to 29% higher).
1.22
Reasons for Leaving Before Graduating
12
10
Frequency
8
6
4
2
0
ol
er
s
k
ic
ic
ng
ie
ea
m
em
om
th
hi
lit
ch
le
br
O
ac
bi
ad
b
on
rs
ro
a
i
Te
ns
Ac
Ec
he
d
po
g/
ee
al
ot
in
on
es
N
an
s
R
vi
rs
d
Ad
Pe
ily
n
te
or
at
Fa
Po
To
The most common reasons that college seniors have for leaving during their senior year before
graduation seem to be for academic purposes or financial reasons rather than personal reasons.
“Economic reasons”, “To attend another school” and “Academic reasons” were the top three reasons
given by 27 of the 42 students interviewed. (about 64%). “Family responsibilities”, “Personal
Problems: or just “needing a break” were reported by just 9 of those interviewed (about 21%).
1.23 a.
Grade Frequency Relative Frequency
A+ 11 0.306
A 10 0.278
B 3 0.083
C 4 0.111
D 4 0.111
F 4 0.111
Total 36 1.0
4
Water quality ratings of California Beaches
0.3
Relativ e Freq.
0.2
0.1
A A+ B C D F
Rating
Assuming that an A+ means little risk of getting sick, most beaches in California seem quite
safe. Two thirds of the beaches are rated a B or higher.
b. No, a dotplot would not be appropriate. “Rating” is categorical data and a dotplot is used for
small numerical data sets.
1.24
Airline Violations
0.5
Relative Frequency
0.4
0.3
0.2
0.1
0
Security Maintenance Flight Hazardous Other
Operations Materials
The two most common violations are security and maintenance. Each is responsible for more than
double the fines imposed by the rest of the violations put together (18%).
1.25 The relative frequencies must sum to 1, so since .40+.22+.07 = .69, it must be that 31% of those
surveyed replied that sleepiness on the job was not a problem.
Sleepy at Work? Relative Frequency

Not at all 0.31
Few days each month 0.40
Few days each week 0.22
Daily Occurrence 0.07
5
Sleepy on the Job?
0.45
0.4
Relative frequency
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
Not at all few days each few days each daily
month week occurrence
1.26
Type of family Percentage
Non-family 29
Married with children 27
Married without children 29
Single parent family 15
TOTAL 100
Make-up of American Suburbs
35
30
25
Percent
20
15
10
5
0
Non-family Married with Married single
children without parent
children family
6
1.27
Most U.S. business schools have acceptance rates of between 16% and 38%. One school has a
slightly lower rate than this (12%) and three schools have a much higher acceptance rate (between
43% and 49%) than the rest of the schools.
1.28
100
90
80
70
Frequency
60
50
40
30
20
10
0
Fame and Money Attract Like Sports Easy Life Don't need an Other
Celebrity Women education
1.29
Sport Frequency Rel.

Freq.
Touch Football (TF) 38 0.226
Soccer (SO) 24 0.143
Basketball (BK) 19 0.113
Baseball/Softball 11 0.065
(BA)
Jogging/Running 11 0.065
(JR)
Bicycling (BI) 11 0.065
Volleyball (VO) 7 0.042
Others (OT) 47 0.280
168 0.999
7
Sport Injuries
0.3
0.25
relative frequency
0.2
0.15
0.1
0.05
0
un
ll
le
l
er
l
er
al
al
Ba
/S
yc
th
cc
/R
b
'b
BA
B'
V'
C
O
TF
So
g
Jo
1.30
Reasons for Nonresponse
0.3
0.
0.2
0.
0.1
0.
0.0
0
Other
Government
protection
surveys
this survey
toward the
Privacy
Don't like
No reason
No time
interested
general
Don't like
Hostility
given
Not
in
1.31 The display suggests that a representative value is around .93. The 20 observations are quite
spread out around this value. There is a gap that separates the 3 smallest and the 3 largest cadence
values from the rest of the data.
8
Chapter 2
Exercises 2.1 – 2.9
2.1 a. It is an observational study, since no treatment was imposed.
b. No. Cause-and-effect conclusions cannot be made from an observational study.
2.2 Children who drink sweet drinks may also eat a lot of sweet foods and high calorie food products.
This would also affect the weight of the subjects.
2.3 Better educated, wealthier, and more active people tend to have better health care and look after
themselves better which may give alternative reasons for reduced risk of heart disease.
2.4 a. This is an observational study. Whether or not the children are raised with two or more
animals is not controlled by the experimenter, presumably it’s the decision of the family.
b. One example is where the family lives. A family that lives in a farmhouse in the middle of the
countryside is more like to have more than two animals (and possible a lower allergy rate)
than a family that lives in a high rise in the middle of a city.
2.5 This is an observational study – the diabetes (the treatment) was not imposed by the experimenter
and so the results were simple being observed. No cause-and-effect conclusion can be made on the
basis on an observational study.
2.6 a. No. There are many confounding variables that could cause a decrease in cravings for fatty
foods watching her show (even if it were true!). Using voluntary response from an
observational study is not a good basis to make a cause and effect conclusion!
b. No. The people that answered the survey on DietSmart.com were not a random sample of
the viewers that watch the Oprah show, and so not even representative of the viewers of the
Oprah show. For this reason, they are certainly not representative of all women in the U.S.
or all women who watch daytime talk shows and any results of the survey cannot be
generalized to either of these groups.
2.7 a. If the definition of affluent Americans was having a household income of over $75,000 and
that the sample was a simple random sample.
b. No. This sample is not representative of all Americans, since only affluent Americans were
included.
2.8 Southerners may have different diets or different life styles compared to the rest of the country. This
is an observational study and so no cause-and-effect conclusion can be made.
2.9 It is possible that other confounding variables may be affecting the study’s conclusion. It could be
that men who eat more cruciferous vegetables are also making a conscious choice about eating
healthier foods. A definitive causal connection cannot be made based on an observational study
alone.
2.10 The first method is as follows. Write the graduates’ names on slips of paper (one name per slip).
Place the slips of paper in a container and thoroughly mix the slips. Then, select 20 slips
(graduates), one at a time without replacement. The graduates whose names are on the slips
constitute the random sample.
9
A second method would be to associate with each graduate a number between 1 and 140. Each
graduate is assigned one number and no two graduates are assigned the same number. Then, use
a random number generator on a computer to select 20 distinct random numbers between 1 and 140.
Graduates whose numbers are selected are included in the sample.
2.11 The names appear on the petition in a list ordered from first to last. Using a computer random
number generator, select 30 random numbers between 1 and 500. Signatures corresponding to the
30 selected numbers constitute the random sample.
2.12 Since each case has a case number, write each case number on a slip of paper; one case number
per slip of paper. Place the 870 slips of paper into a container and thoroughly mix the slips. Then,
select 50 slips (one at a time, without replacement) from the container. The 50 cases whose case
numbers are on the 50 selected slips constitute the random sample of 50 cases.
2.13 Stratified sampling would be worthwhile if, within the resulting strata, the elements are more
homogeneous than the population as a whole.
a. As one's class standing increases, the courses become more technical and therefore the
books become more costly. Also, there might be fewer used books available. Therefore, the
amount spent by freshmen might be more homogeneous than the population as a whole.
The same statement would hold for sophomores, juniors, seniors, and graduate students.
Therefore, stratifying would be worthwhile.
b. The cost of books is definitely dependent on the field of study. The cost for engineering
students would be more homogeneous than the general college population. A similar
statement could be made for other majors. Therefore, stratifying would be worthwhile.
c. There is no reason to believe that the amount spent on books is connected to the first letter in
the last name of a student. Therefore, it is doubtful that stratifying would be worthwhile.
2.14 Both procedures are unbiased procedures. It would be helpful to consider any known patterns in the
variability among trees. Different rows may be exposed to different degrees of light, water, insects,
nutrients, etc. which may affect the sugar content of the fruits. If this is true, then Researcher A’s
method may not produce a sample that is representative of the population. However, if the rows and
trees are homogeneous, then the convenience and ease of implementation of Researcher A’s
method should be considered.
2.15 1. Cluster sampling

2. Stratified random sampling
3. Convenience sampling
4. Simple random sample
5. Systematic sampling
2.16 a. Assign the 3500 full time students with a 4-digit number from 0000 to 3500. Identify the 3000
part time students with a 4-digit number from 3501 to 6500. Using a random number
generator, pick 4-digit numbers, discarding repeats and numbers greater that 6500, until 10
are chosen from each stratum. Identify the students from the names on the list. These
named 20 students will constitute the sample.
b. No – because there are more full time students than part time students and the same number
is being selected from each group, the probability of being chosen is higher if you are a part-
time student.
10
2.17 Using a convenience sample may introduce selection bias. Suppose I wanted to know the average
heights of students in all Stats classes, and used the students in my 10:00 class as my subjects. It
may be that all the basketball players (that are generally tall) are in my 10:00 class because they
have an early practice. Not only would my non-basketball 8:00 students be excluded from being part
of the sample, but resulting average height would over-estimate the average height of my students.
The sample must be representative of the population.
2.18 The sample would consist of people who want to be part of the study or who would want their
opinions known. In both cases, they have attributes that separate them from those who don’t want to
be involved; maybe the study involves something they don’t want to do, or they are uninterested in
the topic under consideration. Those that are part of a voluntary response sample tend not to be
representative of the whole population.
2.19 a. Using a random number generator, create 3 digit numbers. The 3 digits represent the page
number, and the number of correctly defined words can be counted. Suppose we want to
choose 10 pages in the book. Repeat this process until 10 page numbers are selected and
count the number of “words” on each page.
b. As the number of words on a page may be influenced by which topic being discussed – and
therefore which chapter, it makes sense to stratify by chapter. Suppose we want to sample 3
pages in each chapter. For the first chapter and using a random number table, generate a 2
digit number. Use this to select a page within the chapter. Repeat this process twice more
until 3 pages are chosen. Find these page numbers in the chapter and count the number of
defined words on that page. Do this for each chapter until 3 pages from each chapter are
chosen.
c. Choose the 10th page and then every 15th page after that.
d. Randomly choose one Chapter and count the words on each page in that chapter.
e.& f. answers will vary
2.20 Bias has been reduced due to the random sampling, and so the sample of 1000 registered voters is a
good representation of the registered voters in California. The sample will adequately reflect the
population even if it consists of a small fraction of the population.
2.21 In theory, this is a good example of stratified random sample. Randomization is used in selecting the
households in California. Another random technique is used to randomize the available subjects in
each family. However, there are several possible sources of bias. It will eliminate households
without a phone. It may be that the person who has recently had a birthday may be busy or too
grumpy to talk to someone and some-one else (who may not be of voting age) responds to the
questions. Someone may lie and give responses that are not true.
2.22 It was an on-line survey sent to 1260 students which generated 132 replies. The population of
interest is all undergraduates. The fact that the non-response rate was high [1128/1260 (89.5%)] and
the sample excludes all who do not have Internet access are potential sources of bias.
2.23 This was a small sample of people who chose to attend the workshop and were interested in dealing
with the city’s financial aid deficit. They are probably not representative of the population of interest.
2.24 The goal of stratifying is to create homogeneous strata, groups that are more similar to each other in
a meaningful way than the other groups. There is no reason to think that the strata for Schemes 1 or
3 would be more homogeneous than the population as a whole. In Scheme 2 however, the strata
may tend to be more similar with respect to income, which may be related to support for a sales tax
increase.
11
2.25 No, it is a volunteer sample with no random selection. There is no reason to think the sample is
representative of U.S. adults. Volunteers may have had a reason to be worried about mercury
contamination.
2.26 Only 2000 completed surveys were returned. Even if all these 2000 doctors stated they were
frustrated with managed care, that does not mean that over half of the doctors in California SAY they
are frustrated with managed care. It may be that the other 17,000 are so happy, they do not bother to
respond to the questionnaire. The wording of the survey is important; leaving the state may have
nothing to do with a doctors view on managed care. Response bias is present.
2.27 It is possible that different subpopulations used the website and the telephone. The different ways of
responding may have attracted different types of people, and it is possible that neither group is
representative of the town’s residents.
2.28 Increased social isolation being measured by asking the respondents to complete surveys is an
example of a response bias. The subjects may lie and answer the questions in a way that they feel
the researchers would like the questions answered because they are being given free internet and
WebTV service and want to continue the service. Not all U.S adults, even if given free access to all
these services would be willing to spend the time filling out the surveys or want the services in the
first place and so these respondents cannot be representative of all U.S adults. To generalize the
results to the general population is selection bias.
2.29 The best method to use would be to use a simple random sample where every different sample has
an equal chance of being selected. For example, if each student in the school is numbered, you
could use a random number generator to select a random sample of them. you could then find them,
ask their opinion and then assume their views are representative of the whole schools.
2.30 Selection bias. The studies looked at women who had sisters, mothers and grandmothers who
already had breast cancer.
2.31 The response variable is the amount of time it takes a car to exit the parking space.
Extraneous factors that may affect the response variable include: location of the parking spaces, time
of day, and normal driving behaviors of individuals. The factor of interest is whether or not another
car was waiting for the parking space.
One possible design for this study might be to select a frequently used parking space in a parking lot.
Station a data collector at that parking space for given time period (say 2pm-5pm) every day for
several days. The data collector will record the amount of time the driver uses to exit the parking
space and whether another car was waiting for that space. The data collector will also note any other
factors that might affect the amount of time the driver uses to exit such as buckling children into their
car seats. Additionally, this same setup should be replicated at several different locations.
2.32 a. Strength of binding.
b. Type of glue.
c. Number of pages in the book and whether the book is paperback or hardback. Other factors
might include type of paper (rough or glossy), proportion of paper made with recycled paper,
and whether there are tear-out pages included in the book.
12
2.33 a. Randomly select a reasonably large sample of students from your school and also choose 2
comparable IQ tests. Call the tests Test 1 and Test 2. Randomly divide your sample into 2
groups and administer Test 1 to the first group and Test 2 to the second group. Record the
IQ scores for each student. Now gather all the students in the same room and let them listen
to a Mozart piano sonata. After playing the sonata, ask the students to return to their groups.
This time administer Test 2 to the first group and Test 1 to the second group. Again, record
the IQ scores for each student.
b. This design includes direct control for several extraneous variables. Each student in the
sample will take an IQ test before and after listening to the Mozart sonata. The IQ tests are
different but comparable each time. Two IQ tests were used to eliminate the possibility that
retaking the same test may make the test easier the second time and thus higher scores
because of retaking the same test may be confounded with the effects of the piano sonata.
The IQ tests were given such that one group started with Test 1 and the other group started
with Test 2. After listening to the sonata, the groups retook the IQ tests, this time taking the
test they had not previously taken. This design was chosen to eliminate the possibility that
one test is more difficult than the other. All the students in the sample listened to the sonata
in the same room, at the same time and under the same conditions, therefore the factor
‘listening conditions’ can be ruled out as a confounding factor.
c. This design has ‘student’ as a blocking factor. Each student’s IQ test score was recorded
before and after listening to the Mozart sonata. This design should ‘block out’ the effects of
differences in IQ scores across students.
d. By randomly selecting the sample and randomly placing students into 2 groups, we expect to
create ‘equivalent’ experimental groups and minimize biases due to unknown, uncontrolled
factors.
2.34 The researchers must randomize the order of the drinks. For example, if all the cyclists drank
chocolate milk first, and the results showed that they could cycle further with chocolate milk, they
wouldn’t know if it was because chocolate milk was the best or whether it was just that it was early in
the day and they weren’t as tired!
2.35 a. Blocking. Homogeneous groups were formed by height.
b. Direct control. The researchers only used right handed people.
2.36 It is important to create comparable groups. If the subjects were allowed to choose their own groups,
it may be the case that the people with the higher IQ would choose to take the test without
distractions whereas people with the lower IQ might choose to have access to distractions such as
email during the test—not a smart choice!
2.37 So many other factors could have contributed to the difference in the pregnancy rate. The only way
that the difference between the two groups could have been attributed to the program was if the 2
groups were originally formed by dividing all the students completely randomly. This would minimize
the effects of any other factors leaving the program the only big difference between the 2 groups.
2.38 We want the 2 groups to be “equivalent” as possible - groups that are as much alike as possible. We
want to ensure that our experiment doesn’t favor one experimental condition over another, for
example, putting all the communities from the healthier neighborhoods in the same group.
2.39 Yes, blocking on gender is useful for this study because ‘Rate of Talk’ is likely to be different for
males and females.
13
2.40 A placebo treatment if often used to see if there is a psychological response in humans to a
treatment. For instance, to test a new drug to reduce stress, by simply taking a sugar pill (the
placebo) may reduce stress. A control group, made up of people in the same circumstances as
those in the experimental and placebo groups would indicate if there was any other reason that
stress might have been reduced; for instance a big holiday season or an event at work.
2.41 It is necessary to divide into groups that are similar by a certain factor and eliminate any differences
in responses. For examples, if you suspect the reaction to a drug may be different between the
genders, it would be better to block by gender first before randomly dividing into the treatment groups
and the place groups.
2.42 i. Testing a new drug. It would be important to have 2 groups, one taking the new drug and
one taking a “pill” that is identical in every way except with no pharmaceutical benefit – a
sugar pill or placebo. If there is any psychological benefit from just the act of taking a pill, the
patient may feel better just because they are taking the “new” drug.
ii. Testing a new anti-depressant drug vs. exercise. The subject obviously knows which
treatment they are getting. Suppose the experimenter has a vested interest in results and
part of the assessment is subjective. The experimenter may be tempted, in a desire for the
new drug to be successful, rate the patients on the drug higher than the patients on the
exercise regime. The experimenter should not know which treatment the patient is getting.
iii. Testing something where the outcome is subjective – both to the subject and to the
experimenter recording the results – it is imperative that neither know which treatment they
are receiving.
iv. It is impossible to blind by anything obvious – gender, age, race.
2.43 a. It is important to have comparable groups. It may be that people who have a mellow
personality (and includes the physical characteristics of positive attitude, lower blood
pressure, and using fewer laxatives) enjoy art and so if they were allowed to choose groups
they would all go and discuss works of art. The two groups should be equivalent.
b. A control is needed for comparison. It may be the social aspect of meeting other people that
makes the physical changes so a control provides a basis for comparison that allows the
researcher to determine if the discussion of art is really what is affecting the women.
2.44 a. These 16 patients were not a random sample of all heart attack victims and so not
representative.
b. This was an observational study and no cause-and-effect conclusions can be made from an
observational study
c. Randomly select a large number of heart attack victims and randomly assign them to one of
two group: bypass surgery plus cell injections and bypass surgery alone. Compare the heart
efficiency improvement for each group to see if there is a significant difference.
2.45 If either the dog handlers or the experimenters knew which patients had cancer, they might give
physical clues (consciously or unconsciously) or simply be rooting for the dog to be successful.
2.46 a. A placebo group allowed the researchers to see if the psychological effect of taking
something (even if it had no medicinal effect) over a period of 6 months helped prevent
reoccurrence of blocked arteries.
b. To make the two groups as similar as possible to reduce any bias
14
c. No, these patients weren’t randomly sampled from all patients that have undergone surgery
to clear blocked arteries and so the results cannot be generalized.
2.47 a. No, the judges wanted to show that one of Pismo’s restaurants made the best chowder.
b. So that the evaluation is not swayed by personal interest.
2.48 This describes the placebo effect. It shows the importance of eliminating as much as possible, any
subjective view of the success (or lack of) any new medication by the use of a placebo which is
identical in every way to the medication except in pharmaceutical terms.
2.49 a. Randomly divide volunteers into 2 groups of 50, one groups gets PH80 and the other group
gets a placebo nasal spray. They are assessed before and after the treatment and any
improvement in PMS symptoms measured. It would be more accurate if neither the subjects
or the assessors and recorders of the PMS symptoms knew which treatment group each
subject had been assigned.
b. A placebo treatment is needed to see if improvement is due to the PH80 or just a the act of
spraying a liquid (with no medicinal qualities) up your nose that improves the symptoms of
PMS.
c. As irritability is so subjective, double-blinding, as described in a, would be advisable.
2.50 a. There is no such thing as a “filling with no active ingredient” It doesn’t exist!
b. The comparison would have been able to have been made between amalgam filling
treatment vs. resin filling treatment
c. There are usually two reasons to use animals for subjects: either to eliminate non-
psychological influence or to reduce human damage in dangerous medical experiments. It is
possible that the experimenter thought the kidney damage warranted the use of non-humans
in this situation.
Exercises 2.51 – 2.56 Answers will vary
2.57 a. It is an observational study; no treatment was imposed.
b. If a child has attention disorder at a young age, parents who may find life difficult may be
more likely to let the child watch TV so they could have a break. The watching of TV wouldn’t
be causing the disorder, they would be watching TV because of the disorder!
2.58 a. It is an observational study; no treatment was imposed.
b. No. No cause-and-effect can be concluded from an observational study.
2.59 If we are considering all people 12 years of age or older, there are probably more single people that
are young, and more widowed people that are old. It tends to be the young who are at higher risk of
being victims of violent crimes (for instance, staying out late at night). Hence, age could be a
potential confounding variable.
2.60 a. Yes, because of the random assignment of children to treatment.
b. No; children were not randomly selected.
15
2.61 Any survey conducted by mail is subject to selection bias; it eliminates any-one who doesn’t have a
permanent address, any-one who is on vacation or for any other reason doesn’t receive mail. Once
they receive the survey, many people consider surveys as junk mail and only respond to those that
elicit strong feelings - resulting in non-response bias (only a few people reply).
2.62 Firstly, the sample consists of only women and male responses may be different from the women’s
responses. Secondly, the participants are all volunteers and volunteer responses usually differ from
those who choose not to participate. And thirdly, the participants are all from the same university
which may not be representative of the entire nationwide college population.
2.63 Let us evaluate this design by considering each of the basic concepts of designing an experiment.
Replication: Each of the 8 actors was watched on tape by many of the primary care doctors.
Direct Control: The actors wore identical gowns, used identical gestures, were taped from the same
position and used identical scripts.
Blocking: not used
Randomization: The article does not indicate if the 720 doctors were randomly divided into 8 groups
of 90 doctors and each group randomly assigned to watch one of the actors on tape, but it is
reasonable to assume this was done.
This design appears to be good because it employs many of the key concepts in designing an
experiment. One possible improvement would be to randomly select the 720 primary care doctors
from the entire population of primary care doctors. By randomly selecting a sample from the entire
population, we can generalize our results of the study to the whole population. In this study, the
conclusions only apply to this group of 720 doctors.
2.64 Since all the conditions under which the experiment was performed are not given in this problem, it is
possible that there are confounding factors in the experiment. Such factors might be the availability
of cigarettes, the odor of cigarettes in the air, the presence of ashtrays, the availability of food, or
magazines in the room that contain cigarette ads. Any of these factors could explain the craving for
cigarettes. Assuming that the researchers were careful enough to control for these extraneous
factors, the conclusion of the study would appear to be valid.
2.65 a. There are several extraneous variables, which could affect the results of the study. Two of
these are subject variability and trainer variability. The researcher attempted to hold these
variables constant by choosing men of about the same age, weight, body mass and physical
strength and by using the same trainer for both groups. The researcher also included
replication in the study. Ten men received the creatine supplement and 9 received the fake
treatment. Although the article does not say, we hope that the subjects were randomly
divided between the 2 treatments.
b. It is possible that the men might train differently if they knew whether they were receiving
creatine or the placebo. The men who received creatine might have a tendency to work
harder at increasing fat-free mass. So it was necessary to conduct the study as a blinded
study.
c. If the investigator only measured the gain in fat-free mass and was not involved in the
experiment in any other way, then it would not be necessary to make this a double blind
experiment. However, if the investigator had contact with the subjects or the trainer, then it
would be a good idea for this to be a double blind experiment. It would be particularly
important that the trainer was unaware of the treatments assigned to the subjects.
2.66 a. By stratifying by province, information can be obtained about individual provinces as well as
the whole country of Canada. Also, alcohol consumption may differ by province, just like we
expect differences among states in the US.
16
b. Occupation is one socioeconomic factor that could be used for stratification. Alcohol
consumption habits may be different based on a person’s job. For example, a corporate
businessman is likely to have more corporate sponsored social events involving alcohol
consumption than a day care worker. Yearly income is another factor to use for stratification.
Since alcoholic drinks are not free, those people with a high yearly income are likely to be
able to afford the alcoholic drink of their choice.
2.67 a. There are 2 treatments in this experiment – standing or squatting and the response variable
is ‘amount of tip’.
b. There are a number of extraneous factors in this experiment. They include table location and
how busy the place is. Blocking could be used to control for these factors. For instance, one
could separately evaluate the differences in tip received between squatting and standing
during busy hours and during slow hours.
c. Blocking would be essential here for the study to be successful. There are several factors
that could be used for blocking. Time of day, smoking vs non-smoking sections and table
location are a few possibilities. Of course, some level of blocking is already used in this
study with individual waiters/waitresses as blocks.
d. All uncontrolled or unrecorded factors will be confounding variables. An additional
confounding variable is economic status of individual customers.
e. The waiter flips a coin to determine whether he would stand or squat at the table. It is
necessary to randomize the treatments as a strategy for dealing with extraneous variables
not taken into account through direct control or blocking. We count on randomization to
create ‘equivalent’ restaurant customers.
2.68 Divide the 500m square plot into 4 equal size subplots, each measuring 250m x 250m, using two
rows and two columns. Now divide each subplot again into 4 equal size smaller plots, each
measuring 125m x 125m, using the same pattern. The result is, the 500m square plot is divided into
16 subplots with 4 rows of 4 subplots in each row. Now arrange the 4 types of grasslands so that
each type appears in every row and column and in every 2x2 subplot. This is done to allow for
repetition for each treatment (different grasslands). List all possible arrangements such that these
conditions are held, and randomly select one to use in the experiment. Randomization is used in
selecting the type of grassland arrangement for the plot as an effective way to even out the influences
of extraneous variables. A few of the possible confounding variables in this experiment include
exposure to sun or shade, proximity to water, slope of the land or possibly the number of worms in
the soil. This study is an experiment since we are observing how a response variable (nesting)
behaves when one or more factors (grasslands) are changed.
2.69 The response variable for the tile is whether it cracked or not in the firing process. Since two different
firings will not have exactly the same temperature, tiles made from each type of clay should be fired
together. Fifty tiles of each type could be used per firing. Since temperature varies within the oven,
the oven should be divided into sections (blocks) where the temperature is the same within a section,
but perhaps differs between sections. Then, files made from each clay type should be placed within
the sections. The positions of the tiles within a section should be determined in a random fashion.
2.70 There are many possible designs any of which would be a reasonable design. Here is one example.
Assume that the paper is published six days a week and assume that the lender is willing to advertise
for six weeks. For the first week randomly select two days of the week on which advertisement one
will be run. Then select randomly two days from the remaining four on which advertisement two will
be run, and run advertisement three on the remaining two days. Repeat this randomization process
for each week of the experiment. If the newspaper has two sections in which an advertisement can
be placed, then randomly select three of the weeks and place the advertisement in section one, with
the advertisement being run in section two during the remaining three weeks. The randomizations
described should control the extraneous factors such as day of week, section of paper, and daily
fluctuations of interest rates.
17
Chapter 3
Exercises 3.1 to 3.14
3.1
3.2
Frequency of using Swear Words in Conversation
3.3
It is clear in both graphs that most Americans did not move between the years 1995 and 2000, and if they did,
they stayed in the same county. It is also easy to see that the smallest proportion moved abroad. However,
without the labels, it is much easier to see on the bar chart that more households moved within state than to a
different state than by using a pie chart. Comparisons of categories with similar relative frequencies can be
difficult to see with a pie chart.
19
3.4
Change in Tobacco Smoking 1999 to 2001

25
23
22
20
% Admitting to Smoking
17
15
13
10
5
3
0
Year 1999 2001 1999 2001 1999 2001
Grade 7th Grade 9th Grade 11th Grade
It appears that in all three grades the percentage of students that admitted to smoking tobacco was
less in the year 2001 than it was in 1999. However, the reduction was greater in 7th graders (6%
down to 3%) and in 9th graders (17% down to 13%) than in 11th graders which was at the highest of
the three levels in 1999 at 23% and only decreased by one percent to 22% in 2001.
3.5
20
3.6
What can you survive without?

0.5
0.4
0.3
0.2
0.1
0.0
No Maybe Yes No Maybe Yes No Maybe Yes
Personal Computer Cell Phone DVD Player
Nearly half the people surveyed stated that they cannot imagine living without their computer, but the
same number could definitely live without their DVD players (maybe they would play the DVDs on
their computer!). Less than 20% would not want to give up their DVD players but about one third of
the people surveyed would not want to give up their cell phone. It appears that people are far less
willing to give up their computers and their cell phone, but of these two, a higher proportion think they
could live without a cell phone than a computer.
3.7 a.
Percentage of Students passing the Exit Exam by district
90
80
70
Pass Rate as %
60
50
40
30
20
10
0
Year 2002 2003 2004 2002 2003 2004 2002 2003 2004
District San Luis Obispo High School San Luis Obispo County State of California
b. The pass rate increased each year for San Luis Obispo High School and the State of CA
from 2002 to 2004 with a sharp rise in 2004. However, in San Luis Obispo County, there was
a drop in the pass rate in 2003, followed by a sharp increase in 2004 when a pass in the
exam was needed for graduation.
21
3.8 a.
How widespread is corruption in public service?
Which elected officials are MORE ethical?
Hardly anyone,
Not sure, 2% 1%
A small Not sure, 6%
number, 20% Democrats,
Neither, 15% 36%
Almost
everyone, 10%
A lot of people,
28% Both equally,
10%
A moderate
number, 39% Republicans,
33%
b. b.
c. It’s very clear from both graphs that “a moderate number” and “a lot of people” are the two
biggest categories in the corruption question, with the “a small number” category being the
only other large category. It is also clear that the “Democrat” “Republican” and “Neither”
categories are the largest in the ethical question. However, because the answers in the
corruption question are ordered, it can be clearly seen in the segmented bar chart that the
most popular answers are in the middle range of the answers, a fact that is not so obvious in
the pie chart.
3.9 a.
22
b. For both age groups, females are more unfit than the overall age group, and men are less
unfit. However, this difference is much less marked in adolescents who on the whole are
much more unfit than their older counterparts.
3.10 a.
b.
How accurate are your weather forecasts?
60%
50%
40%
30%
20%
10%
0%
Extremely Very Somewhat Not too Not at all Not sure
c. Both graphs clearly show that “Very” and “Somewhat” are the top two categories. However,
as the answer choices have an order, it is easier to see in the bar chart that the popular
answers are in the more favorable end of the categories.
23
3.11 a.
Christmas shopping from Mail Order Catalogues
0.6
Relative Frequency
0.5
0.4
Y2000
0.3
Y2002
0.2
0.1
0
Very likely Somewhat Not too likely Not at all likely
likely
Response
The proportion of Christmas shoppers who are “very likely” or “somewhat likely” to use the
Internet has increased from 2000 to 2002, and the proportion who are “not too likely” or “not
at all likely” has decreased. However it should be noted that the vast majority of Christmas
shoppers (71%) are hesitant to do their Christmas shopping on-line.
b.
Christmas online shopping
0.7
0.6
Relative Frequency
0.5
0.4 Y2000
0.3 Y2002
0.2
0.1
0
Very likely Somewhat Not too likely Not at all likely
likely
Response
The proportion of people who felt strongly about mail order catalogues (“very likely” or “not at
all likely”) increased over the 2 year period while the proportion of those who weren’t too sure
decreased.
24
3.12 a. A bar chart would be a better choice. There are 8 categories which may cause some
confusion in a pie chart.
b.
Reason for Leaving the University
12
10
Frequency
8
6
l
g
s
k
s
oo
s
er
em
m
in
ea
on
ie
ch
th
is
le
lit
br
as
bl
dv
O
bi
ob
rs
ro
re
a
si
ra
pr
he
lp
on
d
ic
ee
oo
ic
ot
na
m
sp
m
an
N
P
no
so
re
de
co
nd
er
ily
ca
P
E
te
m
A
at
Fa c.
3.13 a. To
There are too many categories for this pie chart to be effective.
b.
Roadside Hazard by Object Struck
30
25
Percent
20
15
10
5
0
G e nt
F t
er
il
C t
et e
ge r
or b
il
ba ree
h
le
r
s
a
ra
Ra
ve
c
gn ur
itc
Po
th
Po
Br e B
on en
rd
m
C
T
ul
D
O
nk
ua
id
cr
Em
Si
This is much more effective than the pie chart. It is very easy to compare the small
differences between the objects as hazards on the roads.
3.14 a. The number of people killed in highway work zones has varied between 650 and 850 a year
in a cyclical way. There were two peaks, in 1994 and 1999 when over 800 people were killed
and two troughs, in 1992 and 1997 when less than 700 were killed.
b. A pie chart would not have clearly shown the trend over time.
25
3.15 a. Number of people (12 and older) who have smoked in the last month, by State
1 1333579
2 12568 stem = hundreds (of thousands)
3 0067 leaves = tens (of thousands)
4 5
5 47
6 588
7 34
8 6
9 189
10 24
11 2778
12 12
13 4
14 379
15
16 6
17 8
18
19
20 1
21
22
23 4
24
25
26
27 5
28 67
29
30
31
32
33 6
34
35
36
37
38
39
40 5
41
42
43
44 3
45
46
47
48
49
50
51
52
53
54
55 1
26
b. The distribution is skewed to the right with most of the states having values at the lower end
of the scale. Forty out of the 50 states have less than 1,500,000 people who have smoked in
the past month. There are some outliers at the high end of the distribution.
c. No, it does not indicate that tobacco is necessarily a problem in these states. NY, CA, and
TX are the three most heavily populated states in the United States and even if they have the
same proportion of smokers as others states, they will have a higher number of smokers
because of the greater population.
d. No, it would be better to use the proportion of the population of each state that smoked
during the past month. That way, the population of the state would not affect the result.
3.16
Calorie Content (cal/100ml)of 26 Brands of Light
Beers
1 9 stem: tens
2 23788999 leaf: ones
3 001112233459
4 0123
Most brands of light beers have calorie contents of between 27 and 35 cal/100ml. 4 of the brands
have a lower calorie content (between 19 and 23 cal/100ml) and another four have a higher calorie
content (between 39 and 43 cal/100ml)
3.17
Calorie Content (cal/100ml)of 26 Brands of Light
Beers
1L stem: tens
1H 9 leaf: ones
2L 23
2H 788999
3L 00111223345
3H 9
4L 0123
4H
3.18 a.
% of fully credentialed teachers in CA counties
7 5 stem: tens
8 0334555556788899 leaf: ones
9 0011112234444555555556677777777788888899
10 0
Most counties in California have over 90% of their teachers fully credentialed. Los Angeles
County is the lowest with only 75% and Alpine is the only County where 100% of their
teachers are fully credentialed.
27
b.
% of fully credentialed teachers in CA counties
7L
7H 5 stem: tens
8L 0334 leaf: ones
8H 5555567888899
9L 0011112234444
9H 55555555667777777778888899
10L 0
We can now see that there are only three counties with less then 85% of their teachers fully
credentialed.
3.19 a.
Very Large Urban Areas Large Urban Areas
2 369
8 3 0033589
99 4 0366 stem = tens
1178 5 012355 leaves = units
0379 6
2 7
8
3 9
b. Not necessarily. Philadelphia is a larger urban area than Riverside, CA, but has less extra
travel time. However, overall, taking into account all the urban areas mentioned, or if we
were to calculate the average or typical value for each type of area, then we would find that
on the whole, the larger the urban area, the greater the extra travel time.
3.20 a.
% increase in population 1990 to 2000
0 013444555567888899999
1 00000011234444578 stem: tens
2 00113368 leaf: ones
3 01
4 0
5
6 6
b. 48 of the states have an increase in population of 31% or less, and most of these are under
12%. There are two states that have a much larger %increase: Nevada (66%), and Arizona
(40%)
c.
% increase in population 1990 to 2000
WEST EAST
998880 0 134445555678999
4430 1 0000011244578
83100 2 136
10 3
0 4 stem: tens
5 leaf: ones
6 6
The States that show a large % increase in population are in the West. There are 5 states in
the West (out of 19) that has a % increase greater than the maximum increase in the East.
28
3.21
High School drop out rates 1997-1999 by State
0f 555
0s 666667777777 stem: tens
0* 88888889999999999 leaf: ones
1. 00011111
1t 22223333
1f
1s 77
3.22
a. Relative Frequency Distribution for Drug Plan Costs
Class Interval Frequency Relative Frequency

$0 - <$3 7 .1373
$3 - <$6 3 .0588
$6 - <$9 15 .2941
$9 - <$12 11 .2157
$12- <$15 8 .1569
$15 - <$18 4 .0784
$18 - <$21 3 .0588
TOTAL 51
b.
Histogram of Cost per month (in dollars)

30
25
20
Percent
15
10
0
3 6 9 12 15 18 21
Cost per month (in dollars)
Most of the states have plans in the mid range of cost with a few states with very expensive
plans. More states have plans at the cheaper end than there are states at the expensive end
of the distribution.
c. About 71% of the states have a minimum monthly plan of less than $13.00 a month.
29
3.23 a.
Credit Card Balance (in $) Credit Card Bureau Data)

0.0020
0.0015
Density
0.0010
0.0005
0.0000
0 00 00 00 00 0
5 10 20 70 00
15
Balance
b.
Credit Card Balance (in $) Survey Data
0.0020
0.0015
Density
0.0010
0.0005
0.0000
00 0 0 00 00 00 0
10 50 100 20 30 70 00
15
Balance
c. The results for both the survey and the credit bureau show that most college student debt is
less than $2000. However, the results don’t agree exactly; the credit card bureau had more
students with very large debt than the survey data suggested. Maybe some of the students
with very large debts were reluctant to admit to the scale of their problems!
d. Only 132 out of 1260 students replied, an 89.5% nonresponse rate. It may be that students
with a large debt would be reluctant to respond, so this sample was not representative of all
students.
30
3.24 a.
Number of Frequency Relative
Impairments Frequency
0 100 .4167
1 43 .1792
2 36 .1500
3 17 .0708
4 24 .1000
5 9 .0375
6 11 .0458
n = 240 1.0000
b. .4167 + .1792 + .1500 = .7459
c. 1 − .7459 = .2541
d. .1000 + .0375 + .0458 = .1833
3.25 a.
Average # Cell phone mins used per Month by Men
.60
Relativ e Frequency
.50
.40
.30
.20
.10
0 200 400 600 800
# Mins Used
Most men average between 0 and 200 mins a month. Far fewer average between 400 and
800 mins a month.
b.
Average # Cell phone mins used per Month by Women
.60
.50
Relativ e Frequency
.40
.30
.20
.10
0 200 400 600 800
# Mins used
The distribution for men and women is similar in that most women average between 0 and
200 mins as do men. Fewer women average 600 to 800 minutes than men.
31
c. .18 + .56 = .74
.1
d. .74+ = .79
2
.13*3
e. +.08 =.1775
4
3.26 a. If a commute is longer than 45 minutes, then the time is often rounded to the nearest 15 or 30
minutes.
b
Commute Time Frequency Relative Frequency Density
0 to < 5 5200 .0518 .0104
5 to < 10 18200 .1813 .0363
10 to < 15 19600 .1952 .0390
15 to < 20 15400 .1534 .0390
20 to < 25 13800 .1375 .0307
25 to < 30 5700 .0568 .0275
30 to < 35 10200 .1016 .0114
35 to < 40 2000 .0199 .0203
40 to < 45 2000 .0199 .0040
45 to < 60 4000 .0398 .0027
60 to < 90 2100 .0209 .0007
90 to < 120 2200 .0219 .0007
c.
Most working adults commute between 10 minutes and an hour to work. However, there are
a still a surprising number (look at the area) that commute between one and two hours to
work.
d.
Commute Time Relative Frequency Cumulative Relative Frequency
0 to < 5 .0518 .0518
5 to < 10 .1813 .2331
10 to < 15 .1952 .4283
15 to < 20 .1534 .5817
20 to < 25 .1375 .7192
25 to < 30 .0568 .7760
30 to < 35 .1016 .8776
35 to < 40 .0199 .8975
40 to < 45 .0199 .9174
45 to < 60 .0398 .9572
60 to < 90 .0209 .9781
90 to < 120 .0219 1.0
32
1.0
0.9
0.8
0.7
cum f req.
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0 50 100
time
i 0.93 ii 0.34 iii 17 mins
3.27 a.
0 360, 340, 960, 530, 540, 960, 450, 500, 100, 510, 240, 396
1 280, 240, 050, 000, 320, 250, 120, 850, 670, 890, 419
2 100, 400, 120, 250, 320, 400, 460, 700, 730, 109
3 060, 330, 380, 350, 870, 150, 150
4 390, 770
5 320, 700, 220, 850, 770
stem: thousands digit

leaf: hundreds, tens and ones
The stem and leaf display suggests that a typical or representative value is in the stem 2 row,
perhaps someplace in the 2230 range. The display declines as we move to higher stems
and then rises at stem 5 row. There are no gaps in the display. The shape of the display is
not perfectly symmetric but rather appears to stretch out quite a bit in the direction of low
stems.
b.
Relative
Class Interval Frequency
Frequency
0 - <1000 12 0.2553
1000 - < 2000 11 0.2340
2000 - < 3000 10 0.2128
3000 - < 4000 7 0.1489
4000 - < 5000 2 0.0426
5000 - < 6000 5 0.1064
47 1.0000
33
The upper tail of the histogram stretches farther than the lower tail so the histogram is
positively skewed.
12 + 11 23
c. The proportion of subdivisions with total length less than 2000 is = = .4894 or
47 47
approximately 49%.
10 + 7
The proportion of subdivisions between 2000 and 4000 is = .3617 or approximately
47
36%.
3.28 a. We don’t know the width of the last interval “100,000 or more” and the widths of the intervals
are unequal.
b.
c. Only 42.7% of medical student who have completed their residencies have educational debts
of under $5,000. 41.8% of them have educational debts of between $50,000 and $150,000.
It seems that medical students finishing their residency have either a very large debt or a
relatively small debt with only 15.5% of them having a debt of between $5,000 and $50,000.
3.29 a. If the exam is quite easy, then there would be a large number of high scores with a small
number of low scores. The resulting histogram would be negatively skewed.
b. If the exam were quite hard, then there would be a large number of low scores with a small
number of high scores. The resulting histogram would be positively skewed.
c The students with the better math skills would score high, while those with poor math skills
would score low. This would result in basically two groups and thus the resulting histogram
would be bimodal.
3.30 a. and b.
Classes Frequency Rel. Freq. Cum. Rel. Freq.

0 −< 6 2 .0225 .0225
6 −< 12 10 .1124 .1349
12 −< 18 21 .2360 .3709
18 −< 24 28 .3146 .6855
24 −< 30 22 .2472 .9327
30 −< 36 6 .0674 1.0001 ≈ 1.0
n = 50 1.0000
c. (Rel. Freq. for 12 −< 18) = (Cum. Rel. Freq. for < 18) − (Cum. Rel. Freq. for < 12)
= .3709 − .1349 = .2360.
34
d. i The proportion that had pacemakers that did not malfunction within the first year
equals 1 minus the proportion that had pacemakers that malfunctioned within the first
year (12 months), which is 1 − .1349 = .8651.
ii The proportion that required replacement between one and two years after
implantation is equal to the proportion that had to be replaced within the first 2 years
(24 months) minus the proportion that had to be replaced within the first year (12
months). This is 0.6855 − 0.1349 = 0.5506.
e.
i. Approximately 20 months
ii. Approximately 29 months
3.31 a.
Class Frequency Rel. Freq. Density
Intervals
.15 −< .25 8 .02192 0.2192
.25 −< .35 14 .03836 0.3836
.35 −< .45 28 .07671 0.7671
.45 −< .50 24 .06575 1.3150
.50 −< .55 39 .10685 2.1370
.55 −< .60 51 .13973 2.7946
.60 −< .65 106 .29041 5.8082
.65 −< .70 84 .23014 4.6028
.70 −< .75 11 .03014 0.6028
n = 365 1.00001
35
(8 + 14) 22
b. The proportion of days with a clearness index smaller than .35 is = = .06 , which
365 365
converts to 6%.
(84 + 11) 95
c. The proportion of days with a clearness index of at least .65 is = = .26 , which
365 365
converts to 26%.
3.32 Almost all the differences are positive indicating that the runners slow down. The graph is positively
skewed. A typical difference value is about 150. About .02 of the runners ran the late distance more
quickly that the early distance.
3.33
This histogram is symmetric.
This histogram is positively skewed.
36
This is a bimodal histogram. While it is not perfectly symmetric it is close to being symmetric.
This is a bimodal histogram.
This is a negatively skewed histogram.
37
3.34
Class Intervals I II III IV
100 −< 120 5 5 35 20
120 −< 140 10 7 15 10
140 −< 160 40 10 10 4
160 −< 180 10 15 5 25
180 −< 200 5 33 5 11
a. The histogram associated with frequency set I would be symmetric.
b. The histogram associated with frequency set IV would be bimodal.
c. The histogram associated with frequency set III would be positively skewed.
d. The histogram associated with frequency set II would be negatively skewed.
3.35 a. There are several values that have identical or nearly identical x-values yet different y-
values. Therefore, the value of y is not determined solely by x , but also by various other
factors. There appears to be a tendency for y to decrease as x increases.
b. People with low body weight tend to be small people and it is possible their livers may be
smaller than the liver of an average person. Conversely, people with high weight tend to be
large people and their livers may be larger than the liver of an average person. Therefore,
we would expect the graft weight ratio to be large for low weight people and small for high
weight people.
3.36 In general, there appears to be a positive relationship between drop-out rate and poverty
rate. On average, as the poverty rate increases, the drop-out rate increases.
3.37 a.
There is a tendency for y to increase as x does. That is, larger values of gross efficiency
tend to be associated with larger values of percentage type I fibers (a positive relationship
between the variables).
38
b. There are several observations that have identical x-values yet different y-values (for
example, x6 = x7 = 50, but y 6 = 20.5 and y 7 = 20.1 ). Thus, the value of y is not determined
solely by x, but also by various other factors.
3.38 a.
b.
c. There appears to be a negative relationship between age and emissions. As age increases,
emissions decrease.
d. The two scatterplots have very similar patterns. As engine age increases beyond age = 8
years, there is a definite decrease in emissions in both plots. For age = 0, there are several
different y-values which indicates that emissions are not solely determined by age.
39
3.39 There are several observations that have identical or nearly identical x-values yet different y-values.
Therefore, the value of y is not determined solely by x, but also by various other factors. There
appears to be a general tendency for y to decrease in value as x increases in value. There are two
data points which are far removed from the remaining data points. These two data points have large
x-values and small y-values. Their presence might have an undue influence on a line fit to the data.
3.40 a. There is not a deterministic relationship between x and y. This can be determined by the fact
that there are two data points, (100, 222) and (100, 241), which have the same x-value but
different y-values.
b.
c. There appears to be a tendency for oxygen consumption rate to increase as plasma cortisol
concentration increases.
3.41 a.
% of Households w ith a Computer
50
40
30
%
20
10
1985 1990 1995 2000

Y ear
b. There has been an increase in the ownership of computers over time since 1985. At first the
increase was slow and then from 1995 the increase has been increasing at a more rapid
rate.
40
3.42 a.
Is the Average Home Size Changing with Time?

2400
2200
2000
1800
Home size
1600
1400
1200
1000
1950 1970 1990 2003

Year
b. I would be expect the average home size to be approximately 2500 ft2 by 2010.
3.43 a.
Percentage that thought it was a Good Time to buy a House
90
80 Sep 2003
70 Jun 2003
Mar 1991
Percentage
60
50 Mar 1978
40
30
20
10
0
1977 1982 1987 1992 1997 2002
Date
41
b.
Percentage that thought it was a Good Time or a Bad Time to buy a
House over Time
100
80 Sep 2003
Percentage
Jun 2003
60 Mar 1991 Good Time
Mar 1978 Bad Time
40
Mar 1978 Mar 1991 Sep 2003
20
Jun 2003
0
1977 1982 1987 1992 1997 2002
Date
c. The time series plot gives a better trend over time as it shows a time scale, which the bar
chart does not.
3.44 a.
Average Traffic Fatalities per Week by Day
Variable
200 Mon
Tue
Wed
180 Thurs
Fri
Average Fatalities
Sat
Sun
160
140
120
100
1980 1985 1990 1995 2000

Year
b. There are many more fatalities on a Friday, Saturday and Sunday. Most people work
Monday to Friday and tend to go drive more on at the weekend.
c. There has been a steady decrease in the average number of fatalities between 1980 and
2000, more so on Fridays to Sundays than Mondays to Thursdays. This could be because
most states introduced the compulsory use of front, and then rear seat belts in cars during
this time.
3.45 In both 2001 and 2002 the box office sales dropped in Weeks 2, 6 and in the last two weeks of the
summer. The seasonal peaks occurred during Weeks 4, 9 and 13.
42
3.46
Average Verbal SAT defined by First Language
530 519
520
510
SAT score 500
486
490
480
470 462
460
450
440
430
English English and another A language other than
Language English
3.47
Average Verbal and Math scores for the Language
Groups
530 519 521 521

520 513
510
SAT scores
500 486
490 Verbal SAT
480
470 462 Math SAT Score
460
450
440
430
English English and A language
another other than
Language English
For those students who have English as a first language there is very little difference between the
average Verbal and Math scores. For students who speak English and another language, they do
less well in both subjects, but it is more noticeable in the Verbal scores. Those students who speak a
language other than English as their first language score on average about 55 points less than those
who speak only English. However, their average Math scores are as high as the English only
speaking students.
43
3.48
What is most important for Children?
0.6
0.5
Relative Frequency
0.4
Boys
0.3
Girls
0.2
0.1
0
Grades Popular Sports
3.49
3.50
Other Studios Disney

00012 0L 00001233
59 0H 57
1 1L
56 1H 567
0 2L 02
2H 9
3L Stem: Hundreds
3H Leaves: Tens (truncated)
4L
4H
5L 4
The first impression is that there are many films where the tobacco exposure time is small, both for
Disney films and for those made by other studios. Disney has made more films with longer exposure
to tobacco and there is an obvious outlier in the Disney data.
44
3.51 a.
Av. Transp. Exp. for a British household (in Pounds Sterling)

400
375
Av. Transp. Expenditure 350
325
300
275
250
1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000
Year
Percentage of Household expenses spent on Transportation

17.5
17.0
16.5
Percentage
16.0
15.5
15.0
1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000
Year
b. The time-series plots form part (a) do agree with the statement. It is clear from the first graph
that the actual expenditure has been increasing. Although the percentage of household
expenditure looks volatile, in the 10 years of this study, it has varied from 14.9% to 17.2%,
and is small compared with the increase in actual expenditure.
45
3.52
Gender
0.
Relative
0.
0.
Women
0.
Me
0.
0.
0
Advantage Advantag No Don't
being a being a
Respons
More than half the women think that it is an advantage to be a man and a third think there’s no
advantage for either gender. Very few men (and even fewer women) think that it’s an advantage
being a woman.
3.53 a.
Same Salary for Same Work?
0.8
Relative Frequency
0.7
0.6
0.5
Women
0.4
Me
0.3
0.2
0.1
0
Man earns Woman earns Both earn the Don't know
more more same
b.
46
c. It’s much easier to compare the differences using the comparative bar chart. The categories
are next to each other and any difference is easy to see. To compare the size of the slice
between two pie charts is more difficult.
d. The majority of both men and women think that men earn more money for the same work.
Very few men or women think that a woman earns more. More women than women think
that a man earns more or earns the same as woman.
3.54 a.
Home alone?
Never
Some of the
Time
All of the Time Most of the

Time
b.
Home
100
90
80
All of the
70 Time
60
50
40
Most of the
30 Time
20 Some of the
Time
10
Never
0
c. It is easier to see the difference in the proportions with a pie chart, however it is easier to
estimate the percentage of each response with a segmented bar chart.
47
3.55 a.
Who would you prefer to work for?
No opinion
Prefer to work
for a Man
No difference
Prefer to work
for a Woman
Given the choice, most people don’t mind whether they work for a man or a woman. Of
those that do have an opinion, most would prefer to have a male boss.
b.
Would would you prefer to work for?
100%
90%
80%
No opinion
70%
60% No difference
50%
Prefer to work for a
40% Woman
30% Prefer to work for a Man
20%
10%
0%
3.56 a. 2.2 liters/min.
b. In the row with stem of 8. The leaf of 9 would be placed to the right of the other leaves.
c. A large number of flow rates are between 6.0 and 8.0. Perhaps 6.9 or 7.0 could be selected
as a typical flow rate.
48
d. There appears to be quite a bit of variability in the flow rates. While there are a large
number of flow rates in the 6.0 to 9.0 range, the flow rates appear to vary quite a bit in
relation to one another.
e. The distribution is not symmetric. Taking 7.0 as a typical value, the smaller flow rates are
spread from 2.2 to 7.0, while the larger flow rates are spread from 7.0 to 18.9 (a larger
spread).
f. The value 18.9 appears to be somewhat removed from the rest of the data and hence is an
outlier.
Class Frequency Rel. Freq.

Interval
30 −< 50 2 .04
50 −< 70 4 .08
70 −< 90 4 .08
90 −< 110 11 .22
110 −< 130 12 .24
130 −< 150 3 .06
150 −< 170 5 .10
170 −< 190 4 .08
190 −< 210 3 .06
210 −< 230 1 .02
230 −< 250 1 .02
n = 50 1.00
3.57
49
3.58
How much does a Hurricane Cost?

60
50
40
Cost
30
20
10
0
1990 1992 1994 1996 1998 2000 2002 2004
Year
The peaks coincide with a major hurricane (or several major ones) hitting mainland causing loss of
buildings and life. The top 5 costliest hurricanes were: Katrina (2005), Andrew (1992), Charley
(2004), Ivan (2004) and Wilma (2005).
3.59
0 675890050
1 8615216706312284318087877
2 56907869061971
3 7550001505
4 5678230
5
6 71 stem: tens digit
7 0 leaf: ones digit
The data values are concentrated between 0 to 40, with a few larger values. Overall, the plot
appears to be skewed to the right.
3.60 a.
Day of Week Frequency
Sunday 109
Monday 73
Tuesday 97
Wednesday 95
Thursday 83
Friday 107
Saturday 100
n = 664
50
(107 + 100 + 109 ) 316
b. = = .4759 , which converts to 47.59%.
664 664
c. If a murder were no more likely to be committed on some days than on other days, the
proportion of murders on a specific day would be 1/7 = .1429. So, for three days the
proportion would be 3(.1429) = .4287. Since the proportion for the weekend is .4759, there is
some evidence to suggest that a murder is more likely to be committed on a weekend day
than on a non-weekend day.
3.61 If 39% of those with critical housing needs lived in urban areas and 42% lived in suburban areas,
then 19% (100 – 39 – 42) lived in rural areas.
Area of those with Critical Housing needs
Rural
Urban
Suburban
3.62 a.
Living-donor Transplants by Relative
4000
Number of donors
3500
3000
2500
1994 1995 1996 1997 1998 1999 2000 2001

Year
The number of transplants from a living relative has been increasing steadily from 1994 to
2001.
51
b.
Living-donor Kidney Transplants
4500
4000
3500
3000
2500 related
2000 unrelated
1500
1000
500
0
1994 1995 1996 1997 1998 1999 2000 2001
There are many more related donors than unrelated donors used for kidney transplants.
However the use of non-related donors is rising at a faster rate than that of non-related
donors.
3.63 a. The two histograms do give different impressions about the distribution of values. For the
first histogram, it appears that more frozen meals have sodium content around 800mg.
However, the second histogram suggests that sodium content is fairly uniform from 300mg to
900mg and then drops off above 900mg.
b. Using the first histogram, the proportion of observations that are less than 800 is
approximately
10
6+7+
2 18
= = .6667.
(6 + 7 + 10 + 4) 27
Using the second histogram, the proportion of observations that are less than 800 is
approximately
⎛ 800 − 750 ⎞
6+5+5+⎜ ⎟ (6) 16 + 2 18
⎝ 150 ⎠ = = = .6667.
(6 + 5 + 5 + 6 + 4 + 1) 27 27
18 2
The actual proportion is = = .6667.
27 3
3.64 a.
Class Frequency Rel. Freq.
.175 −< .225 4 .0727
.225 −< .275 2 .0364
.275 −< .325 16 .2909
.325 −< .375 15 .2727
.375 −< .425 9 .1636
.425 −< .475 6 .1091
.475 −< .525 2 .0364
.525 −< .575 0 .0000
.575 −< .625 1 .0182
n = 55 1.0000
52
b. The histogram is skewed slightly in the positive direction. The majority of the observations
are in the center.
53
Chapter 4
9707 707 + 769 1476

4.1 mean = = 970.7 median = = = 738 . There are a couple of outliers at the
10 2 2
high end of the data set and so the mean is influenced by these more than the median. For this
reason, the median is more representative of the sample.
4.2 a. The mean is larger for this data set because there is an outlier at the high end that will
affect the mean, but not the median.
b. Mean = 683315.2, median = 433246.5
c. The median. The distribution is skewed to the right and with the outlier at the high end, the
mean would be much higher than the typical value.
d. The top 20 newspapers in the country is not a random sample of daily U.S. newspapers and
therefore not representative of the population.
4.3 a.
b. Mean = 3.654%, median = 3.35%.
c. An 8% trimmed mean will drop the outlier of 10.1% from the high end of the values. This
will make the 8% trimmed mean smaller than the mean.
4.4 a. Mean = 23.42%
b. It is possible that the lowest score; 14% is from DC. If this is eliminated, the mean is 23.6%,
which rounds to 24%. However, the mean of the data set to the nearest percent is 23%, not
24%.
4.5 a. Mean = 448.3 fatalities.
b. Median = 446 fatalities.
c. The 20 days that had the highest number of fatalities are not a random sample of the 365
days of the year and so cannot be used to generalize to the rest of the year.
4.6 a. It suggests that the distribution for angioplasty wait times is skewed to the left and the
distribution for bypass wait times is skewed to the left.
b. No. The median is equivalent to “50% completed within” and so must be a shorter time than
the “90% completed within”.
55
4.7 a. No, we have no knowledge of the costs of any intermediate plans.
b. The number of plans is not given in the data set.
c. Iowa, Minnesota, Montana, Nebraska, North and South Dakota and Wyoming.
d. Florida.
e. Alaska
f. $9.0925
4.8 a.
b. ii. The mean would be about the same as the median.
c. mean = $9.459, median = $9.489.
d.
e. mean = $72.85, median = $68.61.
4.9 Expensive houses do not pull the median price up. Sales of very expensive houses will increase the
mean price, just as the sales of very inexpensive houses will decrease the mean price. If there are
more sales of either very high or very low-priced homes, this will affect the median price. More
higher-priced sales than lower-priced sales, the median increases; more lower-priced sales, the
median price decreases.
4.10 This statement could be correct if there were a small group of residents with very large wages. This
group would make the average wage very large and thus a large percentage (perhaps even as large
as 65%) could have wages less than the average wage.
4.11 The four observations ordered are: 27.3, 30.5, 33.5, 36.7.
The median = 32 and the mean = 128/4 = 32. If we add the value 32 to the data set, the new ordered
data set ordered is 27.3, 30.5, 32, 33.5, 36.7. For this data set, the sample median is 32 and the
160
mean is x = = 32 .
5
56
7
4.12 a. = .7
10
(1 + 1 + 0 + 1 + 1 + 1 + 0 + 0 + 1 + 1) 7
b. = . They are the same.
10 10
c. There need to be 25(.8) = 20 successes in all. Since there are 7 in the first ten, there need to
be 13 successes in the 15 added observations.
4.13 The median and trimmed mean (trimming percentage of at least 20) can be calculated.
57 + 79 136
sample median = = = 68
2 2
35 + 48 + 57 + 79 + 86 + 92 397
20% trimmed mean = = = 66.17
6 6
4.14 The total for the 19 papers is 19(70) = 1330. To have a class average of 71, the total of all 20 papers
would have to be 20(71) = 1420. Thus, the last paper would have to be a 1420 − 1330 = 90. To have
a class average of 72, the total of all 20 papers would have to be 20(72) = 1440. Thus the last paper
would have to be a 1440 − 1330 = 110. (An impossibility, since the maximum score possible is 100).
6
4.15 n = 5, Σ x = 6, x = = 1.2.
5
x (x - x) (x - x )2
1 −0.2 .04
0 −1.2 1.44
0 −1.2 1.44
3 1.8 3.24
2 0.8 0.64
6 0.0 6.80
6.8
s = = 1.70, s = 1.70 = 1.3038
2
4
4.16 a. Set 1: 2, 3, 7, 11, 12 ; x = 7 and s = 4.528

Set 2: 5, 6, 7, 8, 9 ; x = 7 and s = 1.581
b. Set 1: 2, 3, 4, 5, 6 ; x = 4 and s = 1.581

Set 2: 4, 5, 6, 7, 8 ; x = 6 and s = 1.581
4.17 It would have a large standard deviation. Some parents would spend much more than the average
and some parents would either not be able to afford to spend that much (or not be willing to) or would
rather spread it over the school year.
57
4.18 Standard deviation is a measure of spread and so is volatility in the stock market. A small standard
deviation means the stock stays close to its mean value and a large standard deviation means that it
varies, sometimes it is worth a lot more than its mean, sometimes a lot less, in other words, it is more
volatile.
4.19 a. standard deviation = 50.058 fatalities.
b. Smaller. There is less variability in the data for Memorial Day.
c. There is less variability for holidays that always occur on the same days of the week.
Memorial Day, Labor Day and Thanksgiving Day have standard deviations of 18.2, 17.7 and
15.3 respectively whereas New Year’s Day, July 4th and Christmas Day have standard
deviations of 50.1, 47.1 and 52.4.
4.20 a. i. The lower quartile must be lower than the median of 14.
b. iii The upper quartile is between 13 and 42
c. iv The 95th percentile is greater than 42
4.21 a. The sample size for Los Osos and Morro Bay are different so the average for the combined
( 606456 ) (114) + (511866)(123)
areas is .
(114 + 123)
b. Because Paso Robles has the bigger range of values, it is likely to have the larger standard
deviation.
c. Because the difference in the High prices is so much greater than the difference in the Low
prices, it would suggest that the distribution of house prices in Paso Robles is right skewed
and the mean price is higher than the median suggesting the median would be lower in Paso
Robles than in Grover Beach.
35.07321
4.22 a. s2 = = 3.897 s = 3.897 = 1.974
9
b. Because these were the 10 hospitals with the lowest cost-to-charge ratios, the variability of
the ratios of ALL the hospitals would be greater than for these 10. Therefore, the standard
deviation for all the hospitals in California would be larger than 1.974.
c. The data from these 10 hospitals is, by definition, the lowest cost-to-charge ratio of all the
hospitals in California. They are not representative of all the hospitals in California and so
should not be used to draw conclusions about this population.
20179
4.23 mean = = 747.37 standard deviation = 606.894 ( both in thousands of $)
27
mean + 2 standard deviations = 747.37 +2(606.894) = 1961.158 (or $1,961,158)
4.24 a. Moderately priced midsize cars: standard deviation: $406.98 iqr: $370
b. Inexpensive midsize cars: standard deviation: $186.236 iqr: $155
c. There is more variability in the repair cost for moderately priced midsize cars. Both the
standard deviation and interquartile range values are higher than the corresponding values
for the inexpensive midsize cars.
d. mean repair cost for moderately priced midsize cars: $348.9

mean repair cost for inexpensive midsize cars: $298.36
58
e. On average, the repair cost is higher for a moderately priced midsize car compared to an
inexpensive midsize car. The variability (or spread) of the individual costs is greater for the
moderately priced cars.
4.25 26 data points.

1st quartile will be the middle of the first 13 ordered data points; i.e. the 7th point: 10478
3rd quartile will be the middle of the second 13 ordered data points; i.e. the 20th point: 11778
Interquartile range: 1300
4.26 a. mean: 2965.2 standard deviation: 542.6 iqr: 602
b. As the iqr for the sodium content for chocolate pudding is 602 and the iqr for sodium content
in catsup is 1300, there is less variability in sodium content in chocolate pudding.
4.27 a. 31 data points.

1st quartile will be the middle of the first 15 ordered data points; i.e. the 8th point: 51
3rd quartile will be the middle of the second 15 ordered data points; i.e. the 24th point: 69
Interquartile range: 18
b. The iqr for inpatient cost-to-charge ratio in Example 4.9 is 14. There is more variability in
cost-to-charge ratios for outpatient services.
4.28 a. For sample 1, x = 7.81 and s = .39847

For sample 2, x = 49.68 and s = 1.73897
b. For sample 1, CV = (100)(.39847)/7.81 = 5.10

For sample 2, CV = (100)(1.73897)/49.68 = 3.50
Even though the first sample has a smaller standard deviation than the second, the variation
relative to the mean is greater.
4.29 a. As the mean (22.4) is greater than the median (18), the distribution is positively skewed.
b. Travel Time to Work
0 100 200
c. The maximum point (205) is greater than the upper quartile + 1.5 times the interquartile range
(31 + 1.5(24)). Therefore there is at least one extreme one outlier, possibly more and
possibly some mild outliers.
59
4.30
Nitrous Oxide Emissions (thousands of tons)
0 50 100 150
Half the states in the continental U.S. states have nitrous oxide emissions of less than 34000 tons.
There is one state that has a much larger quantity of emissions than any of the other states The
distribution looks positively skewed meaning that most states have emission values at the lower end
of the range.
4.31 a. mean = 59.85 hours, median = 55.0 hours. As the mean is greater than the median, the
distribution is likely to be skewed to the right.
b.
Extra Travel Time for Commuters
40 50 60 70 80 90 100
Extra Hours
The number of extra commuter hours for urban areas is centered around 55 hours and the
distribution is skewed to the right. There is a major outlier, Los Angeles at 98 hours. Excluding the
outlier the range is from 40 hours to 75 hours.
4.32 a. The iqr because there are two outliers (one low and one high) that would inflate the
standard deviation.
b. The iqr = 94.0-81.5 = 12.5. The mild outliers lie between [upper quartile+1.5(iqr)] and [upper
quartile+3(iqr)] and [lower quartile-1.5(iqr)] and [lower quartile-3(iqr)] or 94+18.75 = 112.75
and 95+37.5 = 132.5 and 81.5-18.75 = 62.75 and 81.5-37.5 = 44. There is one mild outlier at
the lower end; Firefighter (67).
The extreme outliers are greater than [upper quartile+3(iqr)] or less than [lower quartile-
3(iqr)], i.e. > 132.5 or < 44. There are two extreme outliers, one on the high end: Students
(152) and one on the low end: Farmer (43).
60
c.
Number of Auto Accidents per 1000 people for 40 Occupations
50 75 100 125 150

Accidents per 1000 people
d. Apart from authors of statistical text books, it appears that farmers and firemen should be
offered a professional discount as they seem to have far less accidents than other
occupations.
4.33 a. African-American: The mean blood lead level (4.93) is higher than the median (3.6).
This is due to the two unusually large values in the data set that have an
effect on the mean but not on the median.
b.
Blood Lead Level
Whites
African
Americans
0 5 10 15
Overall, African Americans have a higher blood lead level than Whites and also had greater
variability. There were 2 samples from the African Americans that had much higher blood
lead levels than the others of the same ethnicity.
4.34 For Type 1, the ordered data is 655.5, 679.1, 699.4, 721.4, 734.3, 788.3.
699.4 + 721.4
The median = = 710.4
2
The lower quartile is 679.1.
The upper quartile is 734.3.
The interquartile range is 55.2
61
For Type 2, the ordered data is 686.1, 732.1, 772.5, 774.8, 786.9, 789.2.
772.5 + 774.8
The median = = 773.65
2

The interquartile range is 54.8.
717.2 + 727.1
The median is = 722.15
2
The lower quartile is 671.2

542.4 + 559.0
The median is = 550.7
2

3
Type
500 600 700 800

Strength
The median marker for Type 1 is centered in the boxplot whereas the medians for Type 2 and Type 3
are closer to the higher end of their boxplots and the median for Type 4 is closer to the lower end of
its boxplot. The widths of the boxes are approximately equal indicating similar variability in the middle
half of the data for each box type. There are no outliers for any of the different box types.
4.35 a. For the excited delirium sample, the median is 0.4. The lower quartile is 0.1 and the upper
quartile is 2.8. The interquartile range is 2.7.
1.5 + 1.7
For the No Excited Delirium sample, the median is = 1.6. The lower quartile is 0.3
2
and the upper quartile is 7.9. The interquartile range is 7.6.
62
b. To check for outliers for the Excited Delirium sample, we must compute
1.5 x 2.7 = 4.05 and 3 x 2.7 = 8.1
The lower quartile – 4.05 = 0.1 – 4.05 = -3.05
The upper quartile + 4.05 = 2.8 + 4.05 = 6.85
The lower quartile – 8.1 = 0.1 – 8.1 = -8.0
The upper quartile + 8.1 = 2.8 + 8.1 = 10.9
So 8.9 and 9.2 are mild outliers and 11.7 and 21.0 are extreme outliers for the Excited
Delirium sample.
To check for outliers for the No Excited Delirium sample, we must compute
1.5 x 7.6 = 11.4 and 3 x 7.6 = 22.8
The lower quartile –11.4 = 0.3-11.4 = -11.1
The upper quartile + 11.4 = 7.9 + 11.4 = 19.3 There are no outliers for this sample.
c.
No Excited
Delirium
Excited
Delirium
0 10 20
Cocaine Concentration
The median marker for each boxplot is located at the lower end of the boxplot. The boxplot
for the No Excited Delirium is much wider than for Excited Delirium indicating much more
variability in the middle half of the data. Both boxplots have short whiskers on the lower end
and long whiskers on the higher end. The Excited Delirium sample has 2 mild outliers and 2
extreme outliers. There are no outliers for the No Excited Delirium sample.
4.36 a. The number 35 + 5 = 40 is one standard deviation above the mean. The number 35 − 5 = 30
is one standard deviation below the mean. The value 25 is 2 standard deviations below the
mean and 45 is 2 standard deviations above the mean.
b. The value 25 is 2 standard deviations below the mean and 45 is 2 standard deviations above
the mean. Thus, by Chebyshev's rule, at least 75% of the compact discs have playing times
between 25 and 45 minutes.
c. The value 20 is 3 standard deviations below the mean and 50 is 3 standard deviations above
the mean. Thus, by Chebyshev's rule, at least 89% of the playing times are between 20 and
50 minutes. This implies that no more than 11% of the compact discs have playing times
which are either less than 20 minutes or greater than 50 minutes.
63
d. The value 25 is 2 standard deviations below the mean and 45 is 2 standard deviations above
the mean. If the distribution of times is normal, then according to the empirical rule, roughly
95% of the compact discs will have playing times between 25 and 45 minutes. Approximately
.3% will have playing times less than 20 minutes or greater than 50 minutes. Approximately
.15% will have playing times of less than 20 minutes.
4.37 a. The value 57 is one standard deviation above the mean. The value 27 is one standard
deviation below the mean. By the empirical rule, roughly 68% of the vehicle speeds were
between 27 and 57.
b. From part a it is determined that 100% − 68% = 32% were either less than 27 or greater than
57. Because the normal curve is symmetric, this allows us to conclude that half of the 32%
(which is 16%) falls above 57. Therefore, an estimate of the percentage of fatal automobile
accidents that occurred at speeds over 57 mph is 16%.
4.38 a.
Travel Times to Work
0.035
0.030
0.025
0.020
Density
0.015
0.010
0.005
0.000
0 100 200
trav el time
b. Most people take around 20 minutes to get to work, although the time varies from less than 5
minutes to more than 90 minutes. The distribution is positively skewed, meaning very few
people take a long time to get to work.
c. No, it wouldn’t be appropriate to use the Empirical Rule for this distribution. One of the
conditions for the Empirical rule is a normal distribution and this distribution is skewed.
d. According to the empirical rule approximately 95% of the data lies within 2 standard
deviations of the mean. That implies that approximately (100-95)/2 = 2.5% of the data will lie
below the 27 – 2(24) = -21 mins. This is clearly impossible.
27
e. i. 0 mins is standard deviations below the mean. 75 mins is 2 standard deviations
24
1
above the mean. Since 2 is larger, at least (1- 2 ) × 100% = 75% of the data lies
2
between 0 and 75 minutes.
27 20
ii. 0 mins is = 1.125 standard deviations below the mean. 47 mins is = .833
24 24
standard deviations above the mean. Since 1.125 is larger, at least
1
(1- ) × 100% = 21% of the data lies between 0 and 75 minutes.
1.1252
f. Between Times: Chebyshev’s Actual

0 and 75 mins at least 75% Between 93% and 98%
0 and 47 mins at least 21% Between 87% and 93%
64
4.39 a. The value 59.60 is two standard deviations above the mean and the value 14.24 is two
standard deviations below the mean. By Chebyshev's rule the percentage of observations
between 14.24 and 59.60 is at least 75%.
b. The required interval extends from 3 standard deviations below the mean to 3 standard
deviations above the mean. From 36.92 − 3(11.34) = 2.90 to 36.92 + 3(11.34) = 70.94.
c. x − 2s = 24.76 − 2(17.20) = −9.64
In order for the histogram for NO2 concentration to resemble a normal curve, a rather large
percentage of the readings would have to be less than 0 (since x − 25 = −9.64). Clearly, a
reading cannot be negative, hence the histogram cannot have a shape that resembles a
normal curve.
4.40 Observe that annual adult consumption of wine cannot take on negative values. A normal curve has
about 16% of its values smaller than the mean minus one standard deviation. For this data, the mean
minus one standard deviation is 3.15 − 6.09 = −2.94, a negative number. This suggests that the
distribution of annual wine consumption is positively skewed and the Empirical Rule should not be
used to approximate proportions, and in particular the proportion of adults who consume more than
9.24 gallons.
(625 − 475)
4.41 For the first test the student’s z-score is = 1.5 and for the second test it is
100
(45 − 30)
= 1.875. Since the student’s z-score is larger for the second test than for the first test, the
8
student's performance was better on the second exam.
4.42 I would tell her that out of all the people who took the verbal section, 17% scored higher than she did
and 83% scored at or below her score. For the math section, 94% of those who took the exam
scored at or below her score and only 6% scored higher than she did.
4.43 Since the histogram is well approximated by a normal curve, the empirical rule will be used to obtain
answers for parts a − c.
a. Because 2500 is 1 standard deviation below the mean and 3500 is 1 standard deviation
above the mean, about 68% of the sample observations are between 2500 and 3500.
b. Since both 2000 and 4000 are 2 standard deviations from the mean, approximately 95% of
the observations are between 2000 and 4000. Therefore about 5% are outside the interval
from 2000 to 4000.
c. Since 95% of the observations are between 2000 and 4000 and about 68% are between
2500 and 3500, there is about 95 − 68 = 27% between 2000 and 2500 or 3500 and 4000.
Half of those, 27/2 = 13.5%, would be in the region from 2000 to 2500.
d. When applied to a normal curve, Chebyshev's rule is quite conservative. That is, the
percentages in various regions of the normal curve are quite a bit larger than the values given
by Chebyshev's rule.
65
4.44 a.
b. i. The 86th percentile is approximately 21, because the cumulative relative frequency for
the 20 −< 21 class is .86.
ii. The 15th percentile is approximately 18, because the cumulative relative frequency for
the 17 −< 18 class is .15.
iii. The 90th percentile is approximately 21.67, because the cumulative relative frequency
up to 21 is 86%, up to 22 is 92%, and 21.67 is two-thirds of the way across the class of
21 −< 22. The value two-thirds was chosen because 90% is two-thirds of the way
between 86% and 92%.
iv. The 95th percentile will be in the class 25 −< 26. The cumulative relative frequency at
25 is .92. The 95th percentile will be 3/8 of the way across the
(95 − 92) 3
25 −< 26 class because = . Thus, the 95th percentile is 25.375.
8 8
(10 − 2) 8
v. The 10th percentile will be = of the way across the 17 −< 18 class. Hence,
13 13
the value of the 10th percentile is 17.615.
4.45 The recorded weight will be within 1/4 ounces of the true weight if the recorded weight is between
(50.25 − 49.5) (49.75 − 49.5)
49.75 and 50.25 ounces. Now, = 7.5 and = 2.5.
.1 .1
Also, at least 1 − 1/(2.5)2 = 84% of the time the recorded weight will be between 49.25 and 49.75.
This means that the recorded weight will exceed 49.75 no more than 16% of the time. This implies
that the recorded weight will be between 49.75 and 50.25 no more than 16% of the time. That is, the
proportion of the time that the scale showed a weight that was within 1/4 ounce of the true weight of
50 ounces is no more than 0.16.
4.46 a. A z-score of +2.2 indicates that this person scored 2.2 standard deviations above the mean.
Roughly 95% of the scores are within 2 standard deviations of the mean, so this person
scored in the upper 2.5% of the class.
b. A z-score of −.4 indicates that this person scored .4 standard deviations below the mean.
c. A z-score of −1.8 indicates that this person scored 1.8 standard deviations below the mean.
d. A z-score of 1.0 indicates that this person scored 1 standard deviation above the mean.
Roughly 68% of the scores are within 1 standard deviation of the mean. This means that
about (1−.68)/2 = 16% scored higher than the person. Another way of stating this is that this
person scored at the 84th percentile.
66
e. A z-score of 0 indicates that this person scored at the mean. Because a normal curve is
symmetric, the mean is also the median. Thus half of the scores were higher than this
person's score and half were lower. Another way of saying this, is that this person scored at
the 50th percentile.
4.47 Because the number of answers changed from right to wrong cannot be negative and because the
mean is 1.4 and the value of the standard deviation is 1.5, which is larger than the mean, this implies
that the distribution is skewed positively and is not a normal curve. Since
(6 − 1.4)/1.5 = 3.07, by Chebyshev's rule, at most 1/(3.07)2 = 10.6% of those taking the test changed
at least 6 from correct to incorrect.
320 − 450
4.48 a. z= = −1.86
70
475 − 450
b. z= = 0.36
70
420 − 450
c. z= = −0.43
70
610 − 450
d. z= = 2.29
70
4.49 a.
Class Frequency Rel Freq. Cum. Rel.
Freq.
5 −< 10 13 .26 .26
10 −< 15 19 .38 .64
15 −< 20 12 .24 .88
20 −< 25 5 .10 .98
25 −< 30 1 .02 1.00
n = 50 1.00
b. i. The 50th percentile is between 10 and 15 since the cumulative relative frequency at 10
is .26 and at 15 it is .64. The 50th percentile is approximately
10 + 5(50 − 26)/38 = 10 + 3.158 = 13.158.
ii. The 70th percentile is between 15 and 20 since the cumulative relative frequency at 15
is .64 and at 20 it is .88. The 70th percentile is approximately
15 + 5(70 − 64)/24 = 15 + 1.25 = 16.25.
67
iii. The 10th percentile is between 5 and 10 and is approximately
5 + 5(10 − 0)/26 = 5 + 1.923 = 6.923.
iv. The 90th percentile is between 20 and 25 and is approximately

20 + 5(90 − 88)/10 = 20 + 1 = 21.
v. The 40th percentile is between 10 and 15 and is approximately

10 + 5(40 − 26)/38 = 10 + 1.842 = 11.842.
4.50 The z-score associated with the first stimulus value is (4.2 − 6.0)/1.2 = −1.5 and for the second
stimulus value it is (1.8 − 3.6)/.8 = −2.25. Since the z-score associated with the second stimulus
reading is more negative (−2.25 is less than −1.5), you are reacting more quickly (relatively) to the
second stimulus than to the first.
4443 + 4129 8572

4.51 The sample median determines this salary. Its value is = = 4286.
2 2
The mean salary paid in the six counties is
5354 + 5166 + 4443 + 4129 + 2500 + 2220 23812

= = 3968.67.
6 6
Since the mean salary is less than the median salary, the mean salary is less favorable to the
supervisors.
4.52 a. stem : tens leaf : ones

32 55
33 49
34
35 6699
36 34469
37 03345
38 9
39 2347
40 23
41
42 4
The stem and leaf display suggests that the mean and median will be fairly close to each
other. Most values are between 356 and 375 and there are approximately equal amounts of
values larger or smaller than these central values.
9638
b. x= = 370.69
26
369 + 370
median = = 369.5
2
c. The largest value, 424, could be increased by any arbitrary amount without affecting the
sample median. The median is insensitive to outliers. However, if the largest value was
decreased below the sample median, 369.5, then the value of the median would change.
4.53 a. The mean is $1,110,766 which is larger than the median which is $275,000. There are
probably a few baseball players who earn a lot of money (in the millions of dollars) while the
bulk of baseball players earn in the hundreds of thousands of dollars. The outliers have a
greater influence on the mean value making the value larger than the median.
68
b. If the population was all 1995 salaries and all the salaries were used to compute the mean
then the reported mean was the population mean μ.
4.54 The fact that only half the homes in this county cost less than $278,380 is the definition of a median.
The median is not the midpoint of the range. It is the price that 50% of the homes cost less and 50%
of the houses cost more than this price. If there are no houses costing below $300,000 and unless at
least half the homes in this county cost exactly $300,000, Walker is correct; the median will be above
$300,000
532
4.55 a. x= = 48.364
11
b.
Observation Deviation (Deviation)2
62 13.636 185.9405
23 −25.364 643.3325
27 −21.364 456.4205
56 7.636 58.3085
52 3.636 13.2205
34 −14.364 206.3245
42 −6.364 40.5005
40 −8.364 69.9565
68 19.636 385.5725
45 −3.364 11.3165
83 34.636 1199.6525
−0.004 3270.5455
s 2 = (3270.5455)/10 = 327.05 and s = 327.05 = 18.08. s 2 could be interpreted as the

mean squared deviation from the average distance when detection takes place. This is
327.05 squared centimeters. The standard deviation s could be interpreted as the typical
amount by which a distance deviates from the average distance when detection first takes
place. This is 18.08 centimeters.
4.56 Subtracting 10 from each data point yields:
x (x - x) (x - x )2
52 13.636 185.9405
13 −25.364 643.3325
17 −21.364 456.4205
46 7.636 58.3085
42 3.636 13.2205
24 −14.364 206.3245
32 −6.364 40.5005
30 −8.364 69.9565
58 19.636 385.5725
35 −3.364 11.3165
73 34.636 1199.652
422
x= = 38.364
11
These deviations from the mean are identical to the deviations from the mean for the original data set.
This would result in a variance for the new data set that is equal to the variance of the original data
set. In general, adding the same number to each observation has no effect on the variance or
standard deviation.
69
4.57 Multiplying each data point by 10 yields:
x (x - x) (x - x )2
620 136.36364 18595.04132
230 −253.63636 64331.40496
270 −213.63636 45640.49587
560 76.36364 5831.40496
520 36.36364 1322.31405
340 −143.63636 20631.40496
420 −63.63636 4049.58678
400 −83.63636 6995.04132
680 196.36364 38558.67769
450 −33.63636 1131.40496
830 346.36364 119967.76860
0.00004 327054.54545
x = 483.63636
(327054.54545)
s2 = = 32705.4545 s = 32705.4545 = 180.846
10
The standard deviation for the new data set is 10 times larger than the standard deviation for the
original data set.
4.58 a. The ordered data is:

33, 41, 41, 42, 44, 44, 46, 46, 46, 46, 48, 50, 50, 51, 53, 53, 54, 55, 60
The median value is 46.

The upper quartile is 53.
The lower quartile is 44.
The interquartile range is 53-44 = 9.
To check for mild outliers, we must compute 1.5 x 9 = 13.5. Are there any observations
greater that 53 + 13.5 = 66.5? No. Are there any observations less than 44 – 13.5 = 30.5?
No. There are no mild outliers, therefore there are no extreme outliers either.
b.
The median line is somewhat closer to the lower edge of the box suggesting a bit of positive
skewness in the middle half of the data. The lower whisker is slightly longer than the upper
whisker giving the impression of a slight negative skewness in the extremes of the data.
70
4.59 a. median: 20.88 1st Quartile: 18.09 3rd Quartile: 22.20
b. iqr = 22.2 – 18 09 = 4.11

1.5(iqr)= 1.5(4.11) = 6.165 Anything below 18.09 – 6.165 = 11.925 or above
22.2 + 6.165 = 28 365 would be an outlier. There are 2 outliers, 35. 78 and 36.73.
c.
Oxygen capacity (mL/kg/min) for Male middle-aged Runners
15 25 35
oxy gen capacity
Most of the subjects have an oxygen capacity of between 20 and 25 mL/kg/min. There are 2
subjects with much higher oxygen capacities; of over 35 mL/kg/min.
4.60 The data arranged in ascending order is:
7.6, 8.3, 9.3, 9.4, 9.4, 9.7, 10.4, 11.5, 11.9, 15.2, 16.2, 20.4
139.3
x= = 11.61
12
9.7 + 10.4
The sample median equals = 10.05
2
The sample mean is somewhat larger than the sample median because of the outliers 15.2, 16.2 and
20.4. The sample median is more representative of a typical value since it is not influenced by the
mentioned outliers.
2696
4.61 a. x= = 192.57
14
The data arranged in ascending order is: 160, 174, 176, 180, 180, 183, 187, 191, 194, 200,
205, 211, 211, 244.
The median is the average of the two middle values since n = 14, an even number. The
(187 +191)
median value is = 189.
2
b. The median value would remain unchanged, but the mean value would decrease. The new
2656
mean value is = 189.71.
14
c. The trimmed mean =

174 + 176 + 180 + 180 + 183 + 187 + 191 + 194 + 200 + 205 + 211 + 211 2292
= = 191. The
12 12
1
trimming percentage is x 100 = 7.14%
14
71
d. If the largest observation is 204, then the new trimmed mean =
174 + 176 + 180 + 180 + 183 + 187 + 191 + 194 + 200 1665
= = 185
9 9
If the largest observation is 284, the trimmed mean is the same as in c.
4.62 The mean for the sample of 14 values is
13(119.7692) + 159 1557 + 159 1716

= = = 122.57
14 14 14
119 + 119
4.63 The median is = 119.
2
The lower quartile is 87.
The upper quartile is 182.
The interquartile range is 182 – 87 = 95.
To check for outliers, we calculate

1.5 × iqr = 1.5 × 95 = 142.5
3.0 × iqr = 3.0 × 95 = 285
The upper quartile + 1.5 × iqr = 324.5
The lower quartile – 1.5 × iqr = -55.5
The upper quartile + 3 × iqr = 467
The lower quartile – 3 × iqr = -198
Therefore, 511 is an extreme outlier.
0 100 200 300 400 500

Contamination
The median line is not at the center of the box so there is a slight asymmetry in the middle half of the
data. The upper whisker is slightly longer than the lower whisker. There is one extreme outlier, 511.
4.64 n = 20, ∑ x = 18.51, ∑ x 2 = 17.2555
18.51
x= = 0.9255
20
(18.51)2
17.2555 −
s2 = 20 = 17.2555 − 17.1310 = .1245 =.0066
19 19 19
s = .0066 = 0.0809
72
The data arranged in ascending order is:
.78, .81, .81, .85, .85, .86, .92, .92, .93, .93, .93, .93, .95, .95, .96, .96, 1.00, 1.05, 1.06, 1.06
The median value is (.93 + .93)/2 =0.93.
The boxplot for the cadence data is shown below.
The typical cadence is about 0.93 strides per second. There is not a large amount of variability in
strides per second since the standard deviation is .0809. That is, the variability is small relative to the
mean value. The boxplot reveals a bit of skewness in the middle fifty percent of the data. However,
overall the distribution of cadence is reasonably symmetric.
4.65 Budget: lower quartile = (4.0 + 4.1)/2 = 4.05

upper quartile = (7.7 + 7.9)/2 = 7.80
iqr = 7.80 − 4.05 = 3.75
lower mild outlier cutoff = 4.05 − 1.5(3.75) = −1.575
upper mild outlier cutoff = 7.80 + 1.5(3.75) = 13.425
There are no outliers in the budget data.
Midrange : lower quartile = 6.7
upper quartile = 8.1
iqr = 8.1 − 6.7 = 1.4
lower mild outlier cutoff = 6.7 − 1.5(1.4) = 4.6
lower extreme outlier cutoff = 6.7 − 3(1.4) = 2.5
There are no outliers on the high end. There are two outliers on the low end. A mild
outlier of 4 and an extreme outlier of 1.5.
First-class: lower quartile = 6.6

upper quartile = 7.8
iqr = 7.8 − 6.6 = 1.2
lower mild outlier cutoff = 6.6 − 1.5(1.2) = 4.8
lower extreme outlier cutoff = 6.6 − 3(1.2) = 3.0
upper extreme outlier cutoff = 7.8 + 3(1.2) = 11.4
There is an extreme outlier of 1.8 on the low end and a mild outlier of 9.6 on the high
end.
73
Boxplots for the franchise costs of the three types of hotels are given below.
The median franchise cost is about the same for each type of hotel. The variability of franchise cost
differs substantially for the three types. With the exception of a couple of outliers, first-class hotels
vary very little in the cost of franchises, while budget hotels vary a great deal in regard to franchise
costs. Ignoring the outliers, the distribution of franchise costs for first-class hotels is quite symmetric;
mid-range hotels have a distribution of franchise costs which is slightly skewed to the right; while
budget hotels have a distribution of franchise costs which is skewed to the left.
4.66 Smoking Nonsmoking

x 693.364 961.091
median 693.000 947.000
lower quartile 598.000 903.000
upper quartile 767.000 981.000
s2 10787.700 12952.700
s 103.864 113.810
Boxplots for the two groups are below.
The data strongly suggests that milk volume for nonsmoking mothers is greater than that for smoking
mothers.
4.67 Since the mean is larger than the median, this suggests that the distribution of values is positively
skewed or has some outliers with very large values.
74
4.68
x (x - x) (x - x )2
18 −4.15 17.22
18 −4.15 17.22
25 2.85 8.12
19 −3.15 9.92
23 0.85 0.72
20 −2.15 4.62
69 46.85 2194.92
18 −4.15 17.22
21 −1.15 1.32
18 −4.15 17.22
18 −4.15 17.22
20 −2.15 4.62
18 −4.15 17.22
18 −4.15 17.22
20 −2.15 4.62
18 −4.15 17.22
19 −3.15 9.92
28 5.85 34.22
17 −5.15 26.52
18 −4.15 17.22
443 0.00 2454.48
443
a. x= = 22.15
20
s 2 = 2454.48/19 = 129.183 and s = 129.183 = 11.366
b. The 10% trimmed mean is calculated by eliminating the two largest values (69 and 28) and
the two smallest values (17 and 18). The trimmed mean equals 311/16 = 19.4375. It is a
better measure of location for this data set since it eliminates a very large value (69) from the
calculation. It is almost 3 units smaller than the average.
c. The upper quartile is (21 + 20)/2 = 20.5, the lower quartile is (18 + 18)/2 = 18, and the iqr =
20.5 − 18 = 2.5.
d. upper quartile + 1.5(iqr) = 20.5 + 1.5(2.5) = 24.25

upper quartile + 3.0(iqr) = 20.5 + 3(2.5) = 28.00
The values 25 and 28 are mild outliers and 69 is an extreme outlier.
75
e.
4.69 a. The 16th and 84th percentiles are the same distance from the mean but on opposite sides.
Since 80 is 20 units below 100, the 84th percentile is 20 units above 100, which would be at
120.
b. Since 84 − 16 = 68, roughly 68% of the scores are between 80 and 120. Thus, by the
empirical rule, 120 is one standard deviation above the mean, so the standard deviation has
value 20.
c. The value 90 would have a z-score of (90 − 100)/20 = −1/2.
d. The value 140 would be two standard deviations above the mean and thus roughly
.5(5%)=2.5% of the scores would be larger than 140. Hence 140 would be the 97.5
percentile.
e. The value 40 is three standard deviations below the mean, so only about .5(.3%) or .15% of
the scores are below 40. There would not be many scores below 40.
76
Chapter 5
5.1 a. A positive correlation would be expected, since as temperature increases cooling costs
would also increase.
b. A negative correlation would be expected, since as interest rates climb fewer people would
be submitting applications for loans.
c. A positive correlation would be expected, since husbands and wives tend to have jobs in
similar or related classifications. That is, a spouse would be reluctant to take a low-paying
job if the other spouse had a high-paying job.
d. No correlation would be expected, because those people with a particular I.Q. level would
have heights ranging from short to tall.
e. A positive correlation would be expected, since people who are taller tend to have larger feet
and people who are shorter tend to have smaller feet.
f. A weak to moderate positive correlation would be expected. There are some who do well on
both, some who do poorly on both, and some who do well on one but not the other. It is
perhaps the case that those who score similarly on both tests outnumber those who don't.
g. A negative correlation would be expected, since there is a fixed amount of time and as time
spent on homework increases, time in watching television would decrease.
h. No correlation overall, because for small or substantially large amounts of fertilizer yield
would be small.
5.2 The statement is incorrect. The correlation coefficient measures the extent to which x and y are
linearly related. They may have a strong nonlinear relationship and yet have a correlation of zero.
5.3
77
5.4 Association ≠ Causation. For example, it could be age, or the amount they entertain, or even the age
of their children that has a more important effect on their drinking habits rather than the amount they
earn.
Sugar Depression zx zy zxzy

Consumption x Rate y
150 2.3 -1.726 -1.470 2.537
300 3.0 -.369 -.947 .349
350 4.4 .083 .099 .008
375 5.0 .309 .548 .169
390 5.2 .444 .697 .309
480 5.7 1.259 1.071 1.348
4.720
5.5 a.
∑ zxzy 4.720
r= = = 0.944
n -1 5
The correlation is strong and positive.
b. Increasing sugar consumption doesn’t cause or lead to higher rates of depression, it may be
another reason that causes an increase in both. For instance, a high sugar consumption may
indicate a need for comfort food for a reason that also causes depression.
c. These countries may not be representative of any other countries. It may be that only these
countries have a strong positive correlation between sugar consumption and depression rate
and other countries may have a different type of relationship between these factors. It is
therefore not a good idea to generalize these results to other countries.
78
5.6 a.
Inpatient, Outpatient
x y zx zy zxzy
80 62 0.258 0.246 0.064
76 66 0.022 0.663 0.014
75 63 -0.038 0.351 -0.013
62 51 -0.806 -0.900 0.726
100 54 1.441 -0.587 -0.846
100 75 1.441 1.601 2.307
88 65 0.731 0.559 0.409
64 56 -0.688 -0.379 0.261
50 45 -1.516 -1.525 2.312
54 48 -1.279 -1.213 1.552
83 71 0.435 1.184 0.516
∑ zxzy 7.30
r= = = 0.73
n -1 10
75
Outpatient y
65
55
45
50 60 70 80 90 100
Inpatient, x
There appears to be a reasonably strong positive linear relationship between the cost-to-
charge ratio for inpatient and outpatient services at these Oregon hospitals.
b. There is one hospital, Harney District that has a lower outpatient cost-to-charge ratio or
higher inpatient cost-to-charge ratio than the other ten hospitals.
c. If this observation was removed, the remaining points would all be much closer to a line and
so the correlation coefficient would be greater. The relationship would be stronger.
5.7 The correlation coefficient between household debt and corporate debt would be positive, and the
relationship would be strong. As the household debt increases, the corporate debt increases at a
similar rate; this can clearly be seen on the graph by a constant width between the two lines.
5.8 a. r = 0.1178
79
b.
consumer debt
7
6
5.7 5.8 5.9 6.0 6.1 6.2 6.3
houshold debt
The scatterplot supports the correlation coefficient that indicates a very weak, or no linear
relationship between consumer and household debts.
5.9 a. r = 0.335. There is a weak positive relationship between timber sales and the amount of
acres burned in forest fires.
b No. Correlation does not imply a causal relationship.
5.10 a.
Comparing Heart Rate Responses to Two Exercise Tests
200
195
300 yd run
190
185
180
170 175 180 185 190 195

Shuttle
There does not appear to be any linear relationship between peak heart rate during a shuttle
run and peak heart rate during a 300 yard run.
b. n = 10, Σx = 1788, Σy = 1898, ∑x 2

= 320400, Σy 2 = 360628, Σxy = 339472
(1788 )(1898 )
Sxy = 339472 − = 109.6
10
(1788 )2
Sxx = 320400 − = 705.6
10
80
(1898 )2
Syy = 360628 − = 387.6
10
109.6
r = = .2096
705.6 387.6
The value of .2096 suggests at best a very weak linear relationship between the two
variables. The conclusion is consistent with the one of part a.
c. The value of r does not depend on which variable is labeled x. So switching the labels will
not change the value of r .
5.11 a. Since the points tend to be close to the line, it appears that x and y are strongly correlated in
a positive way.
b. An r value of .9366 indicates a strong positive linear relationship between x and y.
c. If x and y were perfectly correlated with r = 1, then each point would lie exactly on a line.
The line would not necessarily have slope 1 and intercept 0.
5.12 a. 80
70
Exam Score
60
50
40
0 10 20
Test Anxiety
There are several observations that have identical x-values yet different y-values. Thus, the
value of y is not determined solely by x, but also by various other factors. There is one data
point that is far removed from the remaining data points. The plot seems to indicate there
may be a tendency for exam scores to decrease as test anxiety increases.
b. There appears to be a linear relationship between x and y. The scatter plot shows a
tendency for y to decrease as x increases. That is, as test anxiety increases, exam scores
decrease. The relationship may be characterized as a moderate negative relationship.
81
c.
x x2 y y2 xy
23 529 43 1849 989
14 196 59 3481 826
14 196 48 2304 672
0 0 77 5929 0
7 49 50 2500 350
20 400 52 2704 1040
20 400 46 2116 920
15 225 51 2601 765
21 441 51 2601 1071
∑ x = 134 ∑ x = 2436
2
∑ y = 477 ∑ y =26085
2
∑ xy = 6633
(134)(134)
Sxx = 2436 - = 2436 – 1995.11 = 440.89
9
(477)(477)
Syy = 26085 - = 26085 – 25281 = 804
9
(134)(477)
Sxy = 6633 - = 6633 – 7102 = -469
9
−469 −469
r = = = -0.7878
440.89 804 (21)(28.35)
r = -0.7878 indicates a moderate negative linear relationship between test anxiety and
exam score.
d. Correlation measures the extent of association, but association does not imply causation.
It is possible that the two variables are not causally related but they may both be related
to a third variable.
(9620)(7436)
27918 −
5.13 r = 2600
(9620)(9620) (7436)(7436)
36168 − 23145 −
2600 2600
27918 − 27513.2 404.8

= = = .3899
574 1878.04 (23.9583)(43.336)
There is a weak positive linear relationship between high school GPA and first-year college GPA.
5.14 No, because x, artist, is not a numerical variable.
5.15 No. An r value of −0.085 indicates an extremely weak relationship between support for
environmental spending and degree of belief in God.
5.16 The sample correlation coefficient would be closest to −.9. This is because there is an almost perfect
negative linear relationship between speed and time required to travel a fixed distance. That is, as
speed increases, time required to traverse the fixed distance decreases.
82
5.17 a.
There is a weak negative association between pollution and the cost of medical care.
b. x = Pollution and y = cost; Σ x = 191.1, Σ x2 = 6184.05 Σ xy = 177807,

Σ y = 5597, n = 6, x = 31.85, y = 932.833
∑ xy − ∑ n ∑ = 177807 −
( x )( y ) (191.1)(5597)
Sxy = = −457.45
6
(∑ x ) 2 2
= ∑x −
(191.1)
Sxx 2
= 6184.05 − = 97.515
n 6
Sxy −457.45
The slope, b = = = −4.69
Sxx 97.515
The intercept, a = y − bx = 932.833 − ( −4.69)(31.85) = 1082.21
The equation: ŷ = 1082.21 - 4.69x
c. The slope is negative, consistent with the description in part a.
d. Yes, it does support the conclusion that elderly people that live in more polluted areas have
higher medical costs, but care must be taken not to state that the pollution causes the high
medical costs – or even the high medical costs causes the pollution!
5.18 a.
90
80
% who would buy lottery ticket
70
60
50
40
30
20
10
0
4 6 8 10 12 14
Grade
There appears to be positive linear relationship between x, the grade level, and y, the
percentage who said they were more likely to purchase a lottery ticket.
83
b. x = Grade and y = %bought; Σ x = 36, Σ x2 = 344 Σ xy = 2318.2,
Σ y = 237.4, n = 4, x = 9, y = 59.35
∑ xy − ∑ n ∑ = 2318.2 − 4 = 181.6
( x )( y ) (36)(237.4)
Sxy =
(∑ x ) 2 2
= ∑x −
(36)
Sxx = 344 − = 20
2
n 4
Sxy 181.6
The slope, b = = = 9.08
Sxx 20
The intercept, a = y − bx = 59.35 − (9.08)(9) = −22.37
The equation: ŷ = -22.37 + 9.08 x
5.19 a. The dependent variable is the number of fruit and vegetable servings per day. The
independent variable is the number of hours of TV viewed per day.
b. Negative, because as the number of hours of TV watched increases, the number of servings
of fruit and vegetables decreases.
5.20 a. For lower values of patient to nurse ratios, nurse job satisfaction might be low because up to
a point, the more patients a nurse has to look after, the more interesting the job would
become. After a certain number, however, the job would get difficult to do well and might get
frustrating. The relationship might be nonlinear.
b. Patient satisfaction is probably related to the amount of attention received. The higher the
patient to nurse ratio, the less personal attention would be received, so the relationship would
be negative.
c. In an ideal world, there would be no relationship between nurse to patient ratio and patient
quality care; it should be excellent, no matter how many patients each nurse has to care for!
However, quality of care probably declines as the number of patients a nurse must care for
increases. The relationship would be negative.
5.21 a.
Head circumference z-score vs. volume of cerebral grey matter
850
Cerebral Grey Matter (ml) 2-5 y
800
750
700
-1 0 1 2 3
Head circumference z-scores
84
b. n = 18, Σx = 24.35, Σy = 13890 Σx 2 = 49.6775, Σy 2 = 10767400, Σxy = 19501.75
x = 1.3528, y = 771.667
(24.35)(13890)
Sxy = 19501.75 − = 711.667
18
(24.35)2
Sxx = 49.6775 − = 16.737
18
138902
Syy = 10767400 − = 48950
18
Sxy 711.677
r = = = .7863
Sxx Syy 16.737 48950
Sxy 711.667
c. The slope, b = = = 42.521
Sxx 16.737
The intercept, a = y − bx = 771.667 − (42.521)(1.3528) = 714.145
The equation: ŷ = 714.145 + 42.521 x
d. When head circumference z-score is 1.8, the predicted volume of grey matter is 790.68 ml.
e. The least-squares line was calculated using values of z-scores of between -0.75 and 2.8 and
therefore is only valid for values in this range. We don’t know if the relationship between
cerebral grey matter and head circumference z-score remains the same outside these values
and so this equation cannot be used for prediction.
5.22 a. The value of the y-intercept of the line ŷ = -147 + 6.175x is –147. The value of the
slope is 6.175. This means that each unit increase in x is associated with an increase
of 6.175 in y, on average. Thus, each 1 cm increase in snout-vent length is associated
with a 6.175 increase in clutch size, on average.
b. This least squares line should not be used to predict clutch size of a salamander with a
snout-vent length of 22 cm because a 22cm snout-vent length is outside the 30-
70cm snout-vent range of the data set by quite a bit.
5.23 a.
There is a moderately strong positive linear relationship between the percentage of public
schools who were at or above the proficient level in math in 4th and 8th grade in the 8 states.
85
b. x = 4th grade and y = 8th grade; Σ x = 140, Σ x2 = 2586 Σ xy = 3497,
Σ y = 188, n = 8, x = 17.5, y = 23.5
∑ xy − ∑ n ∑ = 3497 − 8 = 207
( x )( y ) (140)(188)
Sxy =
(∑ x ) 2 2
= ∑x −
(140)
Sxx 2
= 2586 − = 136
n 8
Sxy 207
The slope, b = = = 1.522
Sxx 136
The intercept, a = y − bx = 23.5 − 1.522(17.5) = −3.135
The equation: ŷ = -3.135 + 1.522x
c. Predicted 8th grade = -3.135 + 1.522(4th grade percent) ⇒ -3.135 + 1.522(14) = 18 (rounded to
nearest integer). This is 2% lower than the actual 8th grade value of 20 for Nevada.
7436 9620
5.24 a. y= = 2.86 x= = 3.7
2600 2600
(9620)(7436)
27918 −
2600 404.8
b= = = 0.7052
(9620)(9620) 574
36168 −
2600
a = 2.86 – 0.7052(3.7) = 0.2508
Therefore, the equation of the least squares regression line is ŷ = 0.2508 + 0.7052x
b. The equation ŷ = 0.2508 + 0.7052x has slope b = 0.7052, so each one unit increase in x
is associated with an increase of 0.7052 in y, on average. This means that, on average,
each 1.0 increase in high school GPA is associated with an increase of 0.7052 in first
year college GPA.
c. For x = 4.0, ŷ = 0.2508 + 0.7052(4.0) = 0.2508 + 2.8208 = 3.0716
5.25 a.
There appears to be a negative linear association between carbonation depth and the
strength of concrete for a sample of core specimens.
86

b. Σ x = 323, Σ x2 = 14339, Σ xy = 3939.9,
Σ y = 130.8, n = 9, x = 35.889, y = 14.533
∑ xy − ∑ n ∑ = 3939.9 −
( x )( y ) (323)(130.8)
Sxy = = −754.367
9
(∑ x ) 2 2
= ∑x −
(323)
Sxx 2
= 14339 − = 2746.889
n 9
Sxy −754.367
The slope, b = = = −0.275
Sxx 2746.889
The intercept, a = y − bx = 14.533 − ( −0.275)(35.889) = 24.40
The equation: ŷ = 24.4 – 0.275x
c. When depth is 25, predicted strength = 24.4 – 0.275(25) = 17.5
d. The least squares line was calculated using values of “depth” of between 8 mm and 65 mm
and therefore is only valid for values in this range. We don’t know if the relationship between
depth and strength remains the same outside these values and so this equation cannot be
used. A depth of 100 mm is clearly outside these values and it would be unreasonable to
use this equation to predict strength.
5.26 It certainly seems that the sooner the paramedics get there, the higher your chances of survival. The
slope of the least squares line is – 9.30, which means that for every extra minute, on average, the
survival rate decreases by 9.30%.
5.27 The slope is the average increase in the y variable for an increase of one unit in the x variable.
Because the home prices (y variable) dropped by an average of $4000 (-4000) for every (1) mile (x
variable) from the Bay area, the slope is -4000/1 = -4000.
5.28 a. r = 0.70 There is a moderately strong positive linear relationship between sale price and
property size.
b. r = -0.333 There is a very weak negative linear relationship (if any!) between sale price and
land/building ratio.
c. I would use size as it has a correlation coefficient ( r) much closer to |1|.
d. Using x = size and y = sale price: Σ x = 16603, Σ x2 = 40097671 Σ xy = 232691.5,

Σ y = 100.5, n = 10, x = 1660.3, y = 10.05
∑ xy − ∑ n ∑ = 232691.5 −
( x )( y) (16603)(100.5)
S xy = = 65831.35
10
(∑ x ) 2 2
=∑x −
(16603)
S xx 2
= 40097671 − = 12531710.1
n 10
S xy 65831.35
The slope, b = = = 0.00525
S xx 12531710.1
The intercept, a = y − bx = 10.05 − 0.00525(1660.3) = 1.333
The equation: ŷ = 1.333 + 0.00525x
87
5.29 a. Σ x = 240, Σ x2 = 6750 Σ xy = 199750,
Σ y = 7250, n = 11, x = 21.818, y = 659.091
∑ xy − ∑ n ∑ = 199750 − 11 = 41568.182
( x )( y) (240)(7250)
S xy =
(∑ x ) 2 2
=∑x −
(240)
S xx 2
= 6750 − = 1513.636
n 11
Sxy 41568.182
The slope, b = = = 27.462
Sxx 1513.636
The equation: ŷ = 59.925 + 27.462x
b. Concentration with 18% bare ground: 59.925 + 27.462(18) = 554 (to nearest integer)
c No, because the data used to obtain the least squares equation was from steeply sloped
plots, so it would not make sense to use it to predict runoff sediment from gradually sloped
plots. You would need to use data from gradually sloped plots to create a least squares
regression equation to predict runoff sediment from gradually sloped plots.
5.30 a. slope = 244.9 intercept = −275.1
b. 244.9
c. y = −275.1 + 244.9(2) = −275.1 + 489.8 = 214.7
d. No. When shell height (x) equals 1, the equation would result in a predicted breaking
strength of −275.1 + 244.9(1) = −30.2. It is impossible for breaking strength to be a negative
value, so the equation results in a predicted value which is not meaningful.
5.31 a.
The graph reveals a moderate linear relationship between x and y.
88
(1368.1)(80.9)
6933.48 −
16 6933.48 − 6917.456 16.0244
b. b= = = = 0.1123
(1368.1)2 117123.85 − 116981.101 142.7494
117123.85 −
16
80.9 ⎛ 1368.1 ⎞
a= − 0.1123 ⎜ ⎟ = 5.0563 − 0.1123(85.5063) = 5.0563 − 9.6024 = −4.5461
16 ⎝ 16 ⎠
c. The change in vital capacity associated with a 1 cm. increase in chest circumference is
.1123.
The change in vital capacity associated with a 10 cm. increase in chest circumference is
10(.1123) = 1.123.
d. yˆ = −4.5461+ .1123(85) = 4.9994
e. No; this is shown by the fact that there are two data points in the data set whose x values are
81.8, but these data points have different y values.
5.32 It is dangerous to use the least squares line to obtain predictions for x-values outside the range of
those contained in the sample, because there is no information in the sample about the relationship
that exists between x and y beyond the range of the data. The relationship may be the same or it
may change substantially. There is no data to support a conclusion either way.
5.33 a. y = 100 + .75(sy)2 = 100 + 1.5(sy). That person's annual sales would be 1.5 standard
deviations above the mean sales of 100.
sy y− y x− x
b. (y − y ) = r (x − x ) , which implies =r . Hence, −1.0 = r(−1.5) implies
sx sy sx
−1.0
r = = .67.
−1.5
5.34 The denominators of b and of r are always positive numbers. The numerator of b and r is
∑ (x − x )(y − y) . Since both b and r have the same numerator and positive denominators, they will
always have the same sign.
89
5.35 a. ŷ = −424.7 + 3.891x
∑(x − x )(cy − c y ) c ∑(x − x )(y − y )

b. Let y' = cy. Then y ′ = cy . b′ = = = cb
∑(x − x ) 2
∑(x − x )2
a′ = c y − cb x = c( y − b x ) = ca
Both the slope and the y intercept are changed by the multiplicative factor c. Thus, the new
least squares line is the original least squares line multiplied by c.
5.36 a.
Σ x = 55, Σ x2 = 385 Σ xy = 1086,

Σ y = 185.6, n = 10, x = 5.5, y = 18.56
∑ xy − ∑ n ∑ = 1086 − 10 = 65.2
( x )( y) (55)(185.6)
S xy =
(∑ x ) 2 2
=∑x −
(55)
S xx 2
= 385 − = 82.5
n 10
Sxy 65.2
The slope, b = = = 0.7903
Sxx 82.5
The equation: ŷ = 14.213 + 0.7903x
The number of transplants has increased steadily over time.
b.
x y ŷ y - ŷ
1 15 15.0036 -.0036
2 15.7 15.7939 -.0939
3 16.1 16.5842 -.4842
4 17.6 17.3745 .2254
5 18.3 18.1648 .1351
6 19.4 18.9552 .4448
7 20 19.7455 .2545
8 20.3 20.5358 -.2358
9 21.4 21.3261 -.0739
10 21.8 22.1164 -.3164
90
There does appear to be curvature in the residual plot which indicates that the relationship
between year and number of transplants may be better described by a curve rather than a
line.
5.37 a.
b.
Yes, there appear to be large residuals, those associated with the x-values of
40, 50 and 60.
91
c.
x y ŷ y - yˆ
40 58 46.5 11.5
50 34 42.0 −8.0
60 32 37.5 −5.5
70 30 33.0 −3.0
80 28 28.5 −0.5
90 27 24.0 3.0
100 22 19.5 2.5
Yes, the residuals for small x-values and large x-values are positive, while the residuals for
the middle x-values are negative.
5.38 a. If the equation of the least squares line is ŷ = 1082.2 − 4.691x :
x y ŷ residuals
30.0 915 941.47 -26.47
31.8 891 933.03 -42.03
32.1 968 931.62 36.38
26.8 972 956.48 15.52
30.4 952 939.59 12.41
40.0 899 894.56 4.44
b. The correlation coefficient, r = -0.581. It indicates a moderately strong negative linear

relationship between pollution and medical cost.
92
c.
It appears that the areas with the high and low pollution have smaller pollution rates have
smaller residuals than the areas with pollution rates in the middle range. This might warrant
further investigation.
d. The observation is influential. With this observation deleted, the equation of the regression
line is ŷ = 974 – 1.35x, which is quite different than the line based on the complete data set.
5.39 a. The equation of the least-squares line is ŷ = 94.33 − 15.388 x .
b.
x y y residual
.106 98 92.6989 5.30112
.193 95 91.3601 3.63988
.511 87 86.4667 0.53326
.527 85 86.2205 -1.22053
1.08 75 77.7110 -2.71096
1.62 72 69.4014 2.59856
1.73 64 67.7088 -3.70876
2.36 55 58.0143 -3.01432
2.72 44 52.4746 -8.47464
3.12 41 46.3194 -5.31945
3.88 37 34.6246 2.37544
4.18 40 30.0082 9.99184
There appears to be a pattern in the plot. It is like the graph of a quadratic equation.
93
5.40 a. No: The least squares line with observation 11 is : ŷ = -1.1 + 1.29x,
without observation 11: ŷ = -17.59 + 1.59x (not a lot of difference in the slope)
b. Yes. With a residual of 100 – 68.56 = 31.44, when se = 12.185, observation 11 can be
considered an outlier.
c. No: The least squares line with observation 5 is : ŷ = -1.1 + 1.29x,

without observation 5: ŷ = 5.26 + 1.16x (not a lot of difference in the slope)
d. No. With a residual of 100 – 95.65 = 4.35, when se = 12.185, observation 5 cannot be
considered an outlier.
5.41 a. r2 = 15.4%
b. r2 = 16%: No, only 16% of the variability in first-year grades can be attributed to an
approximate linear relationship between first-year college grades and SAT II score so this
does not indicate a good predictor.
5.42 a. A value of r 2 = 0.7664 means that 76.64% of the observed variability in clutch size can
be explained by an approximate linear relationship between clutch size and snout vent
length.
b. To find the value of s e , we need the value of SSResid.

We know r 2 = 1- SSResid/SSTo
Solving for SSResid, we get
SSResid = 10266.954
10266.954
se = = 29.25
14 − 2
Thus, a typical amount by which an observation deviates from the least squares line is 29.25.
29.25.
5.43 a. There does appear to be a positive linear relationship between x and y.
100
90
80
70
60
Runoff
50
40
30
20
10
0
0 50 100
Rainfall
94
b. ∑ x = 798 ∑ x 2
= 63040 ∑ y = 643 ∑ y 2
= 41999 ∑ xy = 51232
798 643
x= = 53.2 y = = 42.87
15 15
(798)(643)
51232 −
15 17024.4
b= = = 0.827
(798)(798) 20586.4
63040 −
15
a = 42.87 – 0.827(53.2) = -1.13
yˆ = −1.13 + 0.827 x
c. For x = 80, yˆ = −1.13 + 0.827(80) = 65.03
d. SSResid = ∑y 2
−a ∑ y − b∑ xy
= 41999 – (-1.13)(643) – 0.827(51232)
= 356.726
356.726
se = = 5.238
15 − 2
e.
Rainfall Runoff Residual
5 4 0.99344
12 10 1.20463
14 13 2.55068
17 15 2.06976
23 15 -2.89208
30 25 1.31911
40 27 -4.95062
47 46 8.26057
55 38 -6.35522
67 46 -8.2789
72 53 -5.41376
81 70 4.14348
96 82 3.73888
112 99 7.50731
127 100 -3.89728
95
10
Residual
0
-5
-10
0 20 40 60 80 100 120 140
Rainfall
Yes, the variability of the residuals appears to be increasing with x, indicating that a linear relationship
may not be appropriate.
5.44 a. n = 38 ∑ x = 704 ∑ x2 = 14752 ∑ y = 45.48 ∑ y2 = 55.444 ∑ xy = 829.48
(704)(45.48)
Sxy = 829.48 − = −13.097
38
(704)2
Sxx = 14752 − = 1709.474
38
(45.48)2
Syy = 55.444 − = 1.012
38
−13.097
the slope, b = = −0.00766
1709.474
intercept, a = 1.1968 − (−.00766)(18.526) = 1.339
ŷ = 1.339 – 0.008x
(Sxy )
2
( −13.097 )2
b. r =
2
= = .0992
(Sxx )(Syy ) (1709.474)(1.012)
c. With r 2 close to 0, the linear relationship between perceived stress and telomere length
accounts for a very small proportion of variability in telomere length.
96
5.45 a. ŷ = 766 + .015(9900) = 914.5
residual = 893 − 914.5 = −21.5
b. The typical amount that average SAT score deviates from the least squares line is 53.7.
c. Only about 16% of the observed variation in average SAT scores can be attributed to the
approximate linear relationship between average SAT scores and expenditure per pupil. The
least-squares line does not effectively summarize the relationship between average SAT
scores and expenditure per pupil.
5.46 a. ŷ = −89.09 + .72907(375) = 184.31

residual = 165 − 184.31 = −19.31
2
b. r =.963
5.47 a.
The plot suggests that the least squares line will give fairly accurate predictions. The least
squares equation is y = 5.20683 − .03421x.
b. The summary statistics for the data remaining after the point (143, .3) is deleted are:
n=9 ∑ x = 1060 − 143 = 917 ∑ x2 = 114514 − (143)2 = 94065
∑ y = 15.8 − .3 = 15.5 ∑ y = 27.82 − (.3) = 27.73

2 2
∑ xy = 1601.1 − (143)(.3) = 1558.2
( ∑ x )2 (917)2
∑ x2 − = 94065 − = 94065 − 93432.1111 = 632.8889
n 9
( ∑ x)( ∑ y) (917)(15.5)
∑ xy − = 1558.2 − = 1558.2 − 1579.2778 = −21.0778
n 9
97
−21.0778
b= = − .0333
632.8889
a = 1.7222 − (−.0333)(101.8889) = 1.7222 + 3.3930 = 5.1151
The least squares equation with the point deleted is ŷ = 5.1151 − .0333x. The deletion of this
point does not greatly affect the equation of the line.
c. For the full data set:
(15.8)2
SSTo = 27.82 − = 27.82 − 24.964 = 2.856
10
SSResid = 27.82−5.2068338(15.8) − (−.03421541)(1601.1)
= 27.82 − 82.2680 + 54.7823 = .3343
.3343
r 2 = 1− = 1 − .1171 = .8829
2.856
For the data set with the point (143, .3) deleted:
(15.5)2
SSTo = 27.73 − = 27.73 − 26.6944 = 1.0356
9
SSResid = 27.73 − 5.1151(15.5) − (−.0333)(1558.2) = 27.73 − 79.28405 + 51.8881 = .33405
.33405
r 2 = 1− = 1 − .3226 = .6774
1.0356
The value of r 2 becomes smaller when the point (143, 0.3) is deleted. The reason for
this is that in the equation r 2 = 1 – SSResid/SSTo, the value of SSTo is lowered by
dropping the point (143, 0.3) but the value of SSResid remains about the same.
1235.470
r = 1−
2
5.48 = 0.9512
25321.368
The coefficient of determination reveals that 95.12% of the total variation in hardness of molded
plastic can be explained by the linear relationship between hardness and the amount of time elapsed
since termination of the molding process.
5.49 a. yˆ = 62.9476 − 0.54975(25) = 49.2

residual = 49.2 –70 = -20.8
b. Since the slope of the fitted line is negative, the value of r is the negative square root of
r 2 . So r = - r 2 = − 0.57 = −0.755
SSresid
c. r 2 = 1−
SSTo
SSresid
0.57 = 1 -
2520
Solving for SSResid gives
SSResid = 1083.6
1083.6
se = = 11.64
8
98
5.50 a. Whether se is small or not depends upon the physical setting of the problem. An se of 2 feet
when measuring heights of people would be intolerable, while an se of 2 feet when measuring
distances between planets would be very satisfactory. It is possible for the linear association
between x and y to be such that r2 is large and yet have a value of se that would be
considered large. Consider the following two data sets:
Set 1 Set 2
x y x y
5 14 14 5
6 16 16 15
7 17 17 25
8 18 18 35
9 19 19 45
10 21 21 55
For set 1, r2 = .981 and se = .378. For set 2, r2 = .981 and se = 2.911.
Both sets have a large value for r2, but se for data set 2 is 7.7 times larger than se for data set
1. Hence, it can be argued that data set 2 has a large r2 and a large se.
b. Now consider the data set
x 5 55 15 45 25 35
y 10.004 10.006 10.007 10.008 10.009 10.010
This data set has r2 = .12 and se = .002266. So yes, it is possible for a bivariate data set to
have both r2 and se small.
c. When r2 is large and se is small, then not only has a large proportion of the total variability in y
been explained by the linear association between x and y, but the typical error of prediction is
small.
5.51 a. When r = 0, then se = sy. The least squares line in this case is a horizontal line with intercept
of y .
b. When r is close to 1 in absolute value, then se will be much smaller than sy.
s e = 1 − (.8) (2.5) =.6(2.5) = 1.5

2
c.
d. Letting y denote height at age 6 and x height at age 18, the equation for the least squares
line for predicting height at age 6 from height at age 18 is
⎛ 1.7 ⎞
(height at age 6) = 46 + .8 ⎜ ⎟ [(height at age 18) − 70] = 7.95 + .544(height at age 18)
⎝ 2.5 ⎠
The value of se is 1 − (.8)2 (1.7) = .6(1.7) = 1.02.
99
5.52 a.
Because of the substantial curvature in the plot, a straight line would not provide an effective
summary of the relationship.
b.
The plot of the transformed variables suggests that the relationship could be modeled by a
straight line.
c. The coefficient of determination between y ′ and x ′ is .973. This suggests that a least-
squares line might effectively summarize the relationship between x ′ and y ′ .
d. When x = 35, x ' = 1.54407

yˆ ′ = 2.01780 − 1.05171(1.54407) = 0.393886
yˆ = 100.393886 = 2.47677
e. Yes, this appears to be the case. To see this, predict y using both approaches. Compare the
values of ∑
( y − yˆ )2 for the two methods. The method of part c results in a lower value for
∑ ( y − yˆ )2 .
100
5.53 a.
Fatality Rate vs Age
3.5
% of Drivers killed in Injury Crashes

3.0
2.5
2.0
1.5
1.0
20 30 40 50 60 70 80 90
Age of Driver
1
b. Table suggests moving x up or y down so let y ' = .
y
c.
Age x Fatality Rate y y'=1/y
20 1 1
25 0.99 1.01
30 0.8 1.25
35 0.8 1.25
40 0.75 1.33
45 0.75 1.33
50 0.95 1.05
55 1.05 0.95
60 1.15 0.87
65 1.2 0.83
70 1.3 0.77
75 1.65 0.61
80 2.2 0.45
85 3 0.33
90 3.2 0.31
Transformed Data vs Age

1.4
1.2
Transformed Data
1.0
0.8
0.6
0.4
0.2
20 30 40 50 60 70 80 90
Age
d. The scatterplot suggests there is a good linear transformation.
101
e. Using transformed data:
Σ x = 825, Σ (x)2 = 52375 Σ xy’ = 643.31,
Σ y’ = 13.3603, n = 15, x = 55, y ' = 0..8907
∑ xy '− ∑ n∑ = 12.72 −
( x )( y ') (825)(13.3603)
S xy ' = = −91.5065
15
(∑ x ) 2 2
=∑x −
(825)
S xx 2
= 52375 − = 7000
n 15
Sxy ' −91.5065

The slope, b = = = −0.0131
Sxx 7000
The intercept, a = y − bx = 0.8907 − ( − 0.0131)(55) = 1.6112
1
The equation: yˆ ' = 1.6112 − 0.0131x or = 1.6112 − 0.0131x , where x = Age and
yˆ
1 1
y = fatality rate. When x = 78, = 1.6112 − 0.0131(78) = .5894 , so yˆ = = 1.697.
ŷ .5894
5.54 a.
The plot is curvilinear, not linear.
b. The plot looks like section 3 of figure 5.31, which suggests going down the ladder on y and/or
x. Both x and y are down the ladder.
102
c.
Yes, this plot is straighter than the plot in part a.
d. Since there are x observations whose values are 0, both log(x) and 1/x cannot be employed.
Another transformation that might be helpful in straightening the plot is cube root of x and
cube root of y.
5.55 a. n = 12 ∑ x = 22.4 ∑ y = 303.1 ∑ x 2 = 88.58 ∑y 2

= 12039.27 ∑ xy = 241.29
(22.4)(303.1)
241.29 −
12 −324.50
r= =
2 2 46.767 4383.47
(22.4) (303.1)
88.58 − 12039.27 −
12 12
−324.5
= = −0.717
(6.84)(66.208)
103
b. n=12 ∑ x = 13.5 ∑ y = 55.74 ∑ x 2
= 22.441 ∑y 2
= 303.3626 ∑ xy = 47.7283
(13.5)(55.74)
47.7283 −
12 −14.9792
r= =
(13.5)2 (55.74)2 7.2535 44.4503
22.441 − 303.3626 −
12 12
−14.9792
= −0.835
(2.693)(6.667)
The correlation between x and y is −.835. Since this correlation is larger in absolute
value than the correlation of part a, the transformation appears successful in straightening the
plot.
5.56 a.
From 1990 to 1999 the number of people waiting for a transplant has increased. Each year,
the number of people added to the waiting list increases.
b. Using the transformation y’ where y’ = y :

Σ x = 55, Σ x2 = 385 Σ xy’ = 391.2
Σ y’ = 64.72, n = 10, x = 5.5, y ' = 6.47
∑ xy '− ∑ n∑ = 391.2 − 10 = 35.24

( x )( y ') (55)(64.72)
S xy ' =
(∑ x ) 2 2
=∑x −
(55)
S xx 2
= 385 − = 82.5
n 10
Sxy ' 35.24
The slope, b = = = 0.427
Sxx 82.5
The intercept, a = y '− bx = 6.47 − 0.427(5.5) = 4.12
104
The equation: y’= 4.12 + 0.427x
c. Using the transformed equation: yˆ = 4.12 + 0.427 x, the predicted number of patients
waiting for an organ transplant in 2000 (year 11) is ŷ = 4.12 + 0.427(11) = 8.817 ⇒
y = 2.269. As y is measured in thousands, we predict 2269 patients awaiting transplant
surgery in 2000.
d. We are assuming the relationship between year and the number awaiting transplant stays the
same outside the range of x-values in the given data range. The further from the data range
a prediction is going to be made, the less accurate it may be. 2010 is further away from the
data used to create the least squares line and we don’t know if the relationship between the
two variables is still the same. I would be less confident to make a prediction if the year was
2010.
5.57 a.
The relationship appears non-linear.
b. Σ x = 7.5, Σ x2 = 13.75 Σ xy = 641.05,

Σ y = 370.1, n = 5, x = 1.5, y = 74.02
∑ xy − ∑ n ∑ = 641.05 −
( x )( y) (7.5)(370.1)
S xy = = 85.9
5
(∑ x ) 2 2
=∑x −
(7.5)
S xx 2
= 13.75 − = 2.5
n 5
Sxy 85.9
The slope, b = = 34.36
Sxx 2.5
The equation: ŷ = 22.48 + 34.36x
105
There is a definite curvature in the residual plot confirming the conclusion in part a.
c.
The value of r2 is higher and the size of the residuals are smaller for the log transformation.
d. y = a + b(x’) where x’ = log10(x)

values of x’ are: -0.30102, 0, 0.17609, 0.30103, 0.39794
Σ x’ = .5740, Σ (x’)2 = .3706 Σ x’y = 73.2836,
Σ y = 370.1, n = 5, xʹ = .1148, y = 74.02
∑ xy − ∑ n ∑ = 73.2836 −
( x )( y) (.5740)(370.1)
S xy = = 30.796
5
106
S xx = ∑x 2
−
( ∑ x) 2
= .3706 −
(.574)2
= 0.3047
n 5
Sxy 30.796
The slope, b = = 101.07
Sxx 0.3047
The intercept, a = y − bx = 74.02 − 101.07(.1148) = 62.417
The equation: ŷ = 62.417 + 101.07x’ ⇒ ŷ = 62.417 + 101.07 log (x)
e. When energy of shock (x) = 1.75, predicted success percent to be 62.417 + 101.07(log 1.75)
= 87.0%. When energy of shock is 0.8, the predicted success would be 62.417 + 101.07(log
0.8) = 52.6%
5.58 a.
b.
107
c.
d.
e.
I would recommend using either the transformation of part d or part e.
108
5.59 a.
The plot does appear to have a positive slope, so the scatter plot is compatible with the
"positive association" statement made in the paper.
b.
This transformation does straighten the plot, but it also appears that the variability of y
increases as x increases.
109
c.
The plot appears to be as straight as the plot in b, and has the desirable property that the
variability in y appears to be constant regardless of the value of x.
d.
This plot has curvature opposite of the plot in part a, suggesting that this transformation has
taken us too far along the ladder.
110
5.60
The relationship between age and canal length is not linear, but curvilinear. Transforming to
1/x produces a scatterplot that is much straighter than the plot above.
5.61 Using x = peak intake and y’ = ln(p/1-p): Σ x = 250, Σ x2 = 16500 Σ xy’ = 61.75,
Σ y’ = 6.4958, n = 5, x = 50, y ' = 1.3
∑ xy '− ∑ n∑ = 61.75 −
( x )( y ') (250)(6.4958)
S xy ' = = −263.04
5
(∑ x ) 2 2 S −263.04
=∑x −
(250) xy '
S xx 2
= 16500 − = 4000. The slope, b = = = −0.065876.
n 5 Sxx 4000
The intercept, a = y '− bx = 1.3 − ( −0.06576)(50) = 4.589
The equation: yˆ ' = 4.589 - 0.0659x
Using the values of a and b from the logistic equation, the probability of survival for a hamster with a
e 4.589 −0.0659(40)
peak intake of 40 μg: p = = .876
1 + e 4.589 −0.0659(40)
111
5.62 a.
Probability of Hatching for Low and Mid Elevation Treatments
0.9 Variable
Low
0.8 Mid
0.7
0.6
Probability
0.5
0.4
0.3
0.2
0.1
0.0
0 1 2 3 4 5 6 7 8
Days
The plots have the characteristic “S-shape” of the logistic plot.
b.
p ⎛ p ⎞
Days Proportion y ' = ln ⎜ ⎟
(1 − p ) ⎝1− p ⎠
1 0.75 3 1.098612
2 0.67 2.030303 0.708185
3 0.36 0.5625 -0.57536
4 0.31 0.449275 -0.80012
5 0.14 0.162791 -1.81529
6 0.09 0.098901 -2.31363
7 0.06 0.06383 -2.75154
8 0.07 0.075269 -2.58669
The resulting best fit line is: y ' = a + bx = 1.513 − 0.587 x ,
where y is the proportion of eggs hatched and x = the days of exposure.
The negative slope mean that the value of b < 0, indicating that the curve starts near 1 for
small x values and then decreases as x increases. In other words, the greater the exposure
time, the lower the probability of hatching.
e1.513 −0.587(3)
c. For 3 days: p= = .438 .
1 + e1.513 −0.587(3)
e1.513 −0.587(5)
For 5 days: p= = .194 .
1 + e1.513 −0.587(5)
d. Somewhere between two days (p = .584) and three days (p = .438).
112
5.63 a. It can be seen form the table that as the elevation increases, the Lichen becomes less
common.
b.
p ⎛ p ⎞
Elevation Proportion y ' = ln ⎜ ⎟
(1 − p) ⎝1− p ⎠
400 0.99 99 4.59512
600 0.96 24 3.178054
800 0.75 3 1.098612
1000 0.29 0.408451 -0.89538
1200 0.077 0.083424 -2.48382
1400 0.035 0.036269 -3.31678
1600 0.01 0.010101 -4.59512
The resulting best fit line is: yˆ ' = a + bx = 7.537 − 0.0079 x , where y is the proportion
of plots with lichen and x = elevation.
c. To estimate the proportion of plots of land where the lichen is classified as “common” at an
elevation of 900m:
e7.537 −0.0079(900)
p= = .6052
1 + e7.537 −0.0079(900)
5.64 The table with a row of proportions of mosquitoes killed:
Concentration 0.10 0.15 0.20 0.30 0.50 0.70 0.95

Proportion killed .2083 .25 .4464 .6078 .8298 .9623 .9608
a.
Proportion of Mosquitos Killed with different Concentrations of Pesticide
1.0
0.9
0.8
0.7
Prop. Killed
0.6
0.5
0.4
0.3
0.2
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Conc.
113
b.
p ⎛ p ⎞
Concentration Proportion y ' = ln ⎜ ⎟
(1 − p ) ⎝ 1− p ⎠
0.1 0.2083 0.263105 -1.3352
0.15 0.25 0.333333 -1.09861
0.2 0.4464 0.806358 -0.21523
0.3 0.6078 1.54972 0.438074
0.5 0.8298 4.875441 1.58421
0.7 0.9623 25.5252 3.239666
0.95 0.9608 24.5102 3.19909
The resulting best fit line is: yˆ ' = a + bx = −1.559 + 5.768 x , where y is the proportion
of mosquitoes killed and x = the concentration of pesticide.
The positive slope, b >0, shows that as the concentration of the pesticide increases,
the proportion of the mosquitoes killed also increases.
⎛ p ⎞
c. When the dose kills 50%, p = 0.5, so: ln ⎜ ⎟ = a + bx
⎝ (1 − p ) ⎠
⎛ 0.5 ⎞
ln ⎜ ⎟ = −1.559 + 5.768 x
⎝ (1 − 0.5) ⎠
ln(1) + 1.559 1.559
x= = = 0.270 . About 0.27 g/cc
5.768 5.768
5.65 a.
Proportion failing vs. Load on the Fabric
0.4
0.3
Proportion failing
0.2
0.1
0.0
0 10 20 30 40 50 60 70 80 90
Load
b.
Prop. p ⎛ p ⎞
Load y ' = ln ⎜ ⎟
failing (1 − p ) ⎝ 1− p ⎠
5 0.02 0.020408 -3.89182
15 0.04 0.041667 -3.17805
35 0.2 0.25 -1.38629
50 0.23 0.298701 -1.20831
70 0.32 0.470588 -0.75377
80 0.34 0.515152 -0.66329
90 0.43 0.754386 -0.28185
114
The resulting best fit line is: y ' = a + bx = −3.579 + 0.0397 x , where y is the proportion of
fabrics failing and x = the load applied.
The positive slope, b >0, shows that as the load or forces increases, the proportion of the
fabrics that fail also increases.
e −3.579 + 0.0397(60)
c. When the load is 60, p = = .232 . .232 lbs per sq. in.
1 + e −3.579 +0.0397(60)
⎛ p ⎞
d. When the failure rate is 5%, p = 0.05, so: ln ⎜ ⎟ = a + bx
⎝ (1 − p ) ⎠
⎛ 0.05 ⎞
ln ⎜ ⎟ = −3.579 + 0.0397 x
⎝ (1 − 0.05) ⎠
ln(0.0526) + 3.579
x= = 15.98 .
0.0397
To have less than a 5% chance of a wardrobe malfunction, a maximum force of 15.5 lbs/sq
in. might be suggested.
5.66 a. r = -0.981 There appears to be a strong negative linear relationship between the amount of
catalyst added to a chemical reaction and the resulting reaction time.
b.
There is a definite curvature to the plot. Linear does not seem the best description of this
relationship. This shows the importance of checking, not only the numerical checks of “a
good fit” but also graphical ones too.
5.67 a. ∑ x = 5.92 , ∑ x = 3.8114, ∑ y = 10.47 ,

For this data set, n= 13, 2
∑ y = 9.885699 , ∑ xy = 5.8464 , x = 0.455, y = 0.805

2
(5.92)(10.47)
5.8464 −
13 5.8464 − 4.7679 1.0785
b= = = = 0.9668
(5.92)(5.92) 3.8114 − 2.6959 1.1155
3.8114 −
13
a = 0.805 – 0.9668(0.455) = 0.3651
The least squares regression line is ŷ = 0.3651 + 0.9668x
b. For a value of x = 0.5, ŷ = 0.3651 + 0.9668(0.5) = 0.8485
115
5.68 a. n = 15, ∑ x = 82.82, ∑ y = 12545, ∑ x2 = 459.9784, ∑ y2 = 12734425, ∑ xy = 67703.9
( ∑ x)( ∑ y) (82.82)(12545)
∑ xy − = 67703.9 − = −1561.227
n 15
( ∑ x )2 (82.82)2
∑ x2 − = 459.9784 − = 2.702
n 15
−1561.227
b= = −577.895
2.702
a = 836.33 − (−577.8953)(5.5213) = 4027.083
ŷ = 4027.083 − 577.895 x
b. The b value of −577.895 is the estimate of the average change in myoglobin level associated
with a one unit increase in finishing time.
c. ŷ = 4027.083 − 577.895(8) = −596.077
The least squares equation yields a negative value for the estimated level of myoglobin when
the finishing time is 8h. This is clearly unreasonable since myoglobin level cannot be a
negative value.
5987.16
5.69 r 2 = 1− = 1 − .3439 = .6561
17409.60
So 65.61% of the observed variation in age is explained by a linear relationship between percentage
of root with transparent dentine for premolars and age.
5987.16 5987.16
se2 = = = 176.0929
36 − 2 34
s e = 176.0929 = 13.27
The typical amount by which an observed age deviates from the least squares line of percentage of
root with transparent dentine and age is 13.27.
5.70 a. The least square line is yˆ = 32.08 + 0.5549 x
x y predicted residual
15 23 40.4048 −17.4048
19 52 42.6245 9.3755
31 65 49.2837 15.7163
39 55 53.7231 1.2769
41 32 54.8330 −22.8330
44 60 56.4978 3.5022
47 78 58.1626 19.8374
48 59 58.7175 0.2825
55 61 62.6020 −1.6020
65 60 68.1513 −8.1513
116
b. SSResid = (−17.4048)2 + (9.3755)2 + . . . + (−1.6020)2 + (−8.1513)2
= 302.9271 + 87.9000 + . . . + 2.5664 + 66.4437
= 1635.6833
(545)2
SSTo = 31993 − = 31993 − 29702.5 = 2290.50
10
1635.6833
r 2 = 1− = 1 − .7141 = .2859
2290.5000
c. Only 28.59% of the observed variation in age is explained by the linear relationship between
percent of root transparent dentine and age. Also, se = 14.3, so a typical prediction error is
quite large. The least squares line does not give very accurate predictions.
( ∑ x )2 (22.027)2
5.71 a. ∑ x2 − = 62.600235 − = 62.600235 − 40.43239 = 22.16784
n 12
( ∑ x ∑ y) (22.027)(793)
∑ xy − = 1114.5 − = 1114.5 − 1455.61758 = −341.11758
n 12
−341.11758
b= = −15.38795
22.16784
a = 66.08333 − (−15.38795)(1.83558) = 66.08333 + 28.24586 = 94.32919
The least squares equation is ŷ = 94.33 − 15.388x
( ∑ y )2 (793)2
b. SSTo = ∑ y 2 − = 57939 − = 57939 − 52404.08333 = 5534.91667
n 12
SSResid = 57939 − 94.32919(793) − (−15.38795)(1114.5)

= 57939 − 74803.04767 + 17149.87028 = 285.82261
285.82261
c. r 2 = 1− = 1 − .05164 = .94836 or 94.836%
5534.91667
2 285.82261
d. se = = 28.582261
10
s e = 28.582261 = 5.34624
A typical prediction error would be about 5.35 percent.
e. Since the slope of the fitted line is negative, the value of r is the negative square root of r2.
So r = − r 2 = − .94836 = − .97384 .
117
5.72 a.
118
b. It appears that log x, log y does the best job of producing an approximate linear relationship.
The least-squares equation for predicting y ' = log y from x ' = log x is ŷ' = 1.61867 − .31646 x ' .
When x = 25, x ' = 1.39764
yˆ ′ = 1.61867 − .31646(1.39764) = 1.17628
yˆ = 101.17628 = 15.0064
5.73 a. A value of .11 for r indicates a weak linear relationship between annual raises and teaching
evaluations.
b. r 2 = (.11)2 = .0121
119
5.74 a. The plot does not suggest a linear relationship. However, the one outlier value (51.3, 49.3)
prevents an accurate interpretation.
b. ŷ = −11.37 + 1.0906(40) = 32.254
2
c. The value of r is not very large and the value of se is 4.70, which is large relative to the size
of the y-values in the sample. A straight line is not very effective in summarizing the
relationship.
d. For the new data set, n = 9, ∑ x = 388.8 − 51.3 = 337.5, ∑ y = 310.3 − 49.3 = 261.0
∑ x = 15338.54 − (51.3 ) = 12706.85, ∑ y = 10072.41 − ( 49.3 ) = 7641.92
2 2 2 2
∑ xy = 12306.58 − (51.3 )( 49.3 ) = 9777.49

(∑ x )
2
(337.5)2
∑x 2
−
n
= 12706.85 −
9
= 50.60
∑ xy − ∑ n ∑ = 9777.49 −
( x )( y ) (337.5)(261.0)
= −10.01
9
−10.01 261 ⎛ 337.5 ⎞

b= = −0.1978 , a = − ( −.1978) ⎜ ⎟ = 36.4175
50.60 9 ⎝ 9 ⎠
ŷ = 36.4175 − .1978x
(∑ y )
2
( 261)2 ( −10.01)2
∑ y 2
−
n
= 7641.92 −
9
= 72.92 , r 2 =
( 50.6 )( 72.92 )
= .027
Without the observation (51.3, 49.3) there is very little evidence of a linear relationship
between fire-simulation consumption and treadmill consumption. One would be very hesitant
to use the prediction equation based on the data set including this observation because this
observation is very influential.
5.75 The summary values are: n = 13, ∑ x = 91, ∑ y = 470, ∑x 2

= 819 , ∑y 2
= 19118
∑ xy = 3867
( ∑ x)( ∑ y) ( ∑ x )2 ( ∑ y )2
∑ xy − = 577, ∑ x 2 − = 182, ∑ y 2 − = 2125.6923
n n n
577
a. b= = 3.1703 a = 36.1538 − 3.1703(7) = 13.9617
182
The equation of the estimated regression line is ŷ = 13.9617 + 3.1703 x
120
b.
The plot with the line drawn in suggests that perhaps a simple linear regression model may
not be appropriate. The scatterplot suggests that a curvilinear relationship may exist
between flood depth and damage. The points for small x-values or large x-values are below
the line, while points for x-values in the middle range are above the line.
c. When x = 6.5, ŷ = 13.9617 + 3.1703(6.5) = 34.5687
d. The scatterplot in part b suggests that the value of damage levels off at between 45 and 50
when the depth of flooding is in excess of 10 feet. Using the least squares line to predict
flood damage when x = 18 would yield a very high value for damage and result in a predicted
value far in excess of actual damage. Since x = 18 is outside of the range of x-values for
which data has been collected, we have no information concerning the relationship in the
vicinity of x = 18. All of these reasons suggest that one would not want to use the least
squares line to predict flood damage when depth of flooding is 18 feet.
5.76 a. ∑(x − x )2 = 2, ∑(y − y )2 = 2, ∑(x − x )(y − y ) = 0
0
r= =0
2(2)
b. If y = 1, when x = 6, then r = .509.

(Comment: Any y value greater than .973 will work.)
c. If y = −1, when x = 6, then r = −.509.

(Comment: any y value less than −.973 will work).
121
5.77 a.
( ∑ x )2 ( ∑ y )2 ( ∑ x)( ∑ y)
b. ∑ x2 − = .2157 , ∑ y 2 − = 3.08 , ∑ xy − = 0.474
n n n
.474
b= = 2.1975
.2157
a = 7.6 − (2.1975)(.93286) = 5.550
The least squares line is ŷ = 5.550 + 2.1975 x
c. s x = .2157 / 6 = .1896, s y = 3.08 / 6 = .7165

.474
r = =.5815
6(.1896)(.7165)
This value of r suggests a moderate positive linear relationship between x and y.
122
d.
x rank y rank (x rank)(y rank)
6.5 5.0 32.5

6.5 7.0 45.5
4.0 2.0 8.0
3.0 1.0 3.0
1.0 3.5 3.5
2.0 3.5 7.0
5.0 6.0 30.0
129.5
7 (8)
2
129.5 −
4 17.5
rs = = = .625
7(6)(8) 28
12
This value is very close to the value of r in part c.
5.78 a.
b. Based on the plot in part a and figure 5.34 a transformation going down the ladder on x or y is
suggested. The transformation log(time) will produce a reasonably straight plot.
123
Chapter 6
6.1 A chance experiment is any activity or situation in which there is uncertainty about which of two or
more possible outcomes will result.
Consider tossing a coin two times and observing the outcome of the two tosses. The possible
outcomes are (H, H), (H, T), (T, H), and (T, T), where (H, H) means both tosses resulted in Heads,
(H, T) means the first toss resulted in a Head and the second toss resulted in a Tail, etc. This is an
example of a chance experiment with four possible outcomes.
6.2 The collection of all possible outcomes of a chance experiment is the sample space for the
experiment.
The sample space for the experiment described in Problem 6.1 is {(H, H), (H, T), (T, H), (T, T)}.
6.3 a. Sample space = {(A, A), (A, M), (M, A), (M, M)}.
b.
c. B = {(A, A), (A, M), (M, A)}

C = {(A, M), (M, A)}
D = {(M, M)}
Only D is a simple event.
d. B and C = {(A, M), (M, A)} = C

B or C = {(A, A), (A, M), (M, A)} = B
6.4 a. A = {Head oversize, Prince oversize, Slazenger oversize, Wimbledon oversize, Wilson
oversize}
b. B = {Wimbledon midsize, Wimbledon oversize, Wilson midsize, Wilson oversize}
c. In part b, Not B = {Head midsize, Head oversize, Prince midsize, Prince oversize, Slazenger
midsize, Slazenger oversize}
d. Wimbledon oversize, Wilson midsize, Wilson oversize}
e. B and C = { Wilson midsize, Wilson oversize }
125
f.
Midsize
Head
Oversize
Midsize
Prince
Oversize
Midsize
Slazenger
Oversize
Midsize
Wimbledon
Oversize
Midsize
Wilson
Oversize
6.5 a. # of defective tires # of defective headlights
0 1
1 1
2 1
3 1
4 1
126
b. Ac = {(0,2), (1,2), (2,2), (3,2), (4,2)} where (i,j) means that the number of defective tires is i
and the number of defective headlights is j.
A ∪ B = {(0,0), (1,0), (2,0), (3,0), (4,0), (0,1), (1,1), (2,1), (3,1), (4,1), (0,2), (1,2)} where (i,j)
means that the number of defective tires is i and the number of defective headlights is j.
A ∩ B = {(0,0), (1,0), (0,1), (1,1)}
c. C = {(4,0), (4,1), (4,2)}. A and C are not disjoint because the outcomes (4,0) and (4,1) are in
both A and C. B and C are disjoint because the outcomes in B are of the form (i,j) where i =0
or 1 but the outcomes in C are of the form (i,j) where i=4. Recall that the notation (i,j) means
that the number of defective tires is i and the number of defective headlights is j.
6.6 a. First book selected Second book selected
1 3
2 3
3 2
4 2
b. A = {(1,2), (1,4), (2,1), (2,3), (2,4), (3,2), (3,4), (4,1), (4,2), (4,3), (4,4)} where (i,j) means that
copy i was selected for the 2-hour service and copy j was selected for the overnight loan.
c. B = {(1,3), (1,4), (2,3), (2,4), (3,1), (3,2), (4,1), (4,2)} where (i,j) means that
copy i was selected for the 2-hour service and copy j was selected for the overnight loan.
127
6.7 a. First book Second book Third book
selected selected selected
2 4
1 3
1 4
2 3
b. Exactly one book is examined when the first selected book is copy 3, 4, or 5. So A = {3,4,5}.
c. C = {5, (1,5), (2,5), (1,2,5), (2,1,5)}. The notation used is as follows. The outcome 5 means
that copy 5 is selected first. The outcome (1,5) means that copy 1 is selected first and copy 5
is selected second, and so on.
6.8 a. N, DN, DDN, DDDN, DDDDN are five possible outcomes. Here DDN means that the first
battery selected is defective, the second battery selected is defective, and the third battery
selected is not defective, etc.
b. The number of outcomes in the sample space is infinite.
c. E = {DN, DDDN, DDDDDN, ......... }. This event consists of all outcomes in which the first n
batteries selected are all defective with n being an odd number, followed by a nondefective
battery.
128
6.9 a. A = {NN, DNN, NDN}
b. B = {DDNN, DNDN, NDDN}
c. The number of possible outcomes is infinite.
6.10 a. The 27 possible outcomes are (1,1,1), (1,1,2), (1,1,3), (1,2,1), (1,2,2), (1,2,3), (1,3,1), (1,3,2),
(1,3,3), (2,1,1), (2,1,2), (2,1,3), (2,2,1), (2,2,2), (2,2,3), (2,3,1), (2,3,2), (2,3,3), (3,1,1), (3,1,2),
(3,1,3), (3,2,1), (3,2,2), (3,2,3), (3,3,1), (3,3,2), (3,3,3).
b. A = {(1,1,1), (2,2,2), (3,3,3)}
c. B = {(1,2,3), (1,3,2), (2,1,3), (2,3,1), (3,1,2), (3,2,1)}
d. C = {(1,1,1), (1,1,3), (1,3,1), (1,3,3), (3,1,1), (3,1,3), (3,3,1), (3,3,3)}
e. Bc = {(1,1,1), (1,1,2), (1,1,3), (1,2,1), (1,2,2), (1,3,1), (1,3,3), (2,1,1), (2,1,2), (2,2,1), (2,2,2),
(2,2,3), (2,3,2), (2,3,3), (3,1,1), (3,1,3), (3,2,2), (3,2,3), (3,3,1), (3,3,2), (3,3,3)}
Cc = {(1,1,2), (1,2,1), (1,2,2), (1,2,3), (1,3,2), (2,1,1), (2,1,2), (2,1,3), (2,2,1), (2,2,2), (2,2,3),
(2,3,1), (2,3,2), (2,3,3), (3,1,2), (3,2,1), (3,2,2), (3,2,3), (3,3,2)}
A ∪ B = {(1,1,1), (2,2,2), (3,3,3), (1,2,3), (1,3,2), (2,1,3), (2,3,1), (3,1,2), (3,2,1)}

A ∩ B = the empty set. There are no outcomes common to A and B.
A ∩ C = {(1,1,1), (3,3,3)}
6.11 a.
E1
E2
E3
129
b. The required region is the area common to all three circles.
E1
E2
E3
c.
E1
E2
E3
130
d.
E1
E2
E3
e.
E1
E2
E3
131
f.
E1
E2
E3
6.12 a. ( A ∪ B) c
A
B
Ac ∩ B c
A
B
We see that ( A ∪ B ) c = Ac ∩ B c .
132
b. ( A ∩ B) c
A B
Ac ∪ B c
A B
We see that ( A ∩ B ) c = Ac ∪ B c .
6.13 a. A = {(C,N), (N,C), (N,N)}. So P(A) = 0.09 + 0.09 + 0.01 = 0.19.
b. B = {(C,C), (N,N)}. So P(B) = 0.81 + 0.01 = 0.82.
1 1 1
6.14 a. P(picking 9, then 1, then another 1) = • • = 0.001
10 10 10
b. Classical or relative frequency.
6.15 a. In the long run, 1% of all people who suffer cardiac arrest in New York City survive.
b. 1% of 2329 is .01(2329) = 23.29. So in this study there must have been only 23 or 24
survivors.
6.16 The article sounds as though you will fly safely for 26,000 years and crash on the very last day.
Obviously no-one lives that long. In fact, there is a very small risk with each flight.
133
500,000
6.17 a. P(twins) = = .0119
42,005,100
100
b. P(quadruplets) = = .00000238
42,005,100
505,100
c. P(more than a single child) = = .012
42,005,100
6.18 a. In the long run, 35% of all customers who purchase a tennis racket at this store will have a
grip size equal to 4 ½ inches.
b. P(grip size is not 4 ½ inches) = 1 - 0.35 = 0.65.
c. P(oversize head) = 0.20 + 0.15 + 0.20 = 0.55. In the long run, 55% of all customers who
purchase a racket at this store will buy a racket with an oversize head.
d. P(grip size is at least 4 ½ inches) = 0.20 + 0.15 + 0.15 + 0.20 = 0.70.
6.19 a. P(owns shares in balanced fund) = 0.07.
b. P(owns shares in a bond fund) = 0.15 + 0.10 + 0.05 = 0.30.
c. P(does not own shares in a stock fund) = 1 - P(owns shares in a stock fund)
= 1 - (0.18 + 0.25) = 0.57.
6.20 a. P(request is for one of the three B’s) = 0.05 + 0.26 + 0.09 = 0.40.
b. P(request is not for one of the two S’s) = 1 - P(request is for one of the two S’s)
= 1 - (0.12+0.07) = 0.81.
c. P(request is for a composer who wrote at least one symphony)

= 1 - P(request is not for Bach or Wagner) = 1-(0.05 + 0.01) = 0.94.
6.21 No, it is not true. The addition rule is for disjoint events. The events “midsize” and “4 3/8 inch grip” are
not disjoint since P(midsize AND 4 3/8 inch grip) = 0.10 is greater than zero.
1287
6.22 a. P(hand will consist entirely of spades) = = 0.000495198 .
2598960
P(hand will consist entirely of a single suit) = P(all spades) + P(all clubs) + P(all diamonds)
+ P(all hearts) = 0.000495198 + 0.000495198 + 0.000495198 + 0.000495198 =
0.001980792.
b. P(hand consists entirely of spades and clubs with both suits represented)
63206
= = 0.0243197.
2598960
c. P(hand contains cards from exactly two suits) = P(spades and clubs) + P(spades and
diamonds) + P(spades and hearts) + P(clubs and diamonds) + P(clubs and hearts) +
P(diamonds and hearts) = 6(0.0243197) = 0.1459182.
134
6.23 a. The 24 outcomes (including the outcome (2, 4, 3, 1) ) are
(1, 2, 3, 4), (1, 2, 4, 3), (1, 3, 2, 4), (1, 3, 4, 2), (1, 4, 2, 3), (1, 4, 3, 2)
(2, 1, 3, 4), (2, 1, 4, 3), (2, 3, 1, 4), (2, 3, 4, 1), (2, 4, 1, 3), (2, 4, 3, 1)
(3, 1, 2, 4), (3, 1, 4, 2), (3, 2, 1, 4), (3, 2, 4, 1), (3, 4, 1, 2), (3, 4, 2, 1)
(4, 1, 2, 3), (4, 1, 3, 2), (4, 2, 1, 3), (4, 2, 3, 1), (4, 3, 1, 2), (4, 3, 2, 1)
b. The required outcomes are

(1, 2, 4, 3), (1, 4, 3, 2), (1, 3, 2, 4), (4, 2, 3, 1), (3, 2, 1, 4), (2, 1, 3, 4).
6
P(exactly two of the 4 books will be returned to the correct owner) = = 0.25 .
24
c. The outcomes corresponding to the event “exactly one students receives his/her book” are
(1, 3, 4, 2), (1, 4, 2, 3), (2, 3, 1, 4), (2, 4, 3, 1), (3, 1, 2, 4), (3, 2, 1, 4), (4, 1, 3, 2), (4, 2, 1, 3).
8
Hence the probability of this event = = 0.33333.
24
d. If three students receive their books, then the fourth student must receive his/her book also.
So, the required probability is 0.
e. The outcomes corresponding to this event are (1, 2, 3, 4), (1, 2, 4, 3), (1, 3, 2, 4), (1, 4, 3, 2),
7
(2, 1, 3, 4), (3, 2, 1, 4), (4, 2, 3, 1). Hence the required probability is = 0.291667.
24
6.24 a. The simple events in this experiment are

(C, D, P), (C, P, D), (D, C, P), (D, P, C), (P, C, D), (P, D, C).
All six outcomes may be considered equally likely so the probability assigned to any one
outcome is 1/6.
b. P(C is ranked first) = 2/6 = 1/3.
c. P(C is ranked first and D is ranked last) = 1/6.
6.25 a. Suppose the students selected are from Mathematics and Physics. We abbreviate this
outcome by {M, P}. With this notation, the ten possible outcomes are {B, C}, {B, M}, {B, P},
{B, S}, {C, M}, {C, P}, {C, S}, {M, P}, {M, S}, {P, S}.
b. Probability of each simple event = 1/10 = 0.1.
c. P(one member is from Statistics) = 4/10 = 0.4.
d. Assuming that the laboratory sciences are Biology, Chemistry, and Physics, the required
probability is 3/10 = 0.3.
6.26 a. The four simple events are (brand M, 2 heads), (brand M, 4 heads), (brand Q, 2 heads),
(brand Q, 4 heads).
b. P(brand Q with 2 heads) = 32% = 0.32.
c. P(brand M) = 25% + 16% = 41% = 0.41.
6.27 a. P(O1) + P(O2) + P(O3) + P(O4) + P(O5) + P(O6) = 1 implies that p + 2p + p + 2p + p + 2p =1.
So. 9p = 1 which yields p = 1/9. The probabilities of the six simple events are as follows:
P(O1) = P(O3) = P(O5) = 1/9
P(O2) = P(O4) = P(O6) = 2/9.
135
b. P(odd number) = P(1 or 3 or 5 ) = 3(1/9) = 3/9 = 1/3 = 0.3333.
P(at most 3) = P(1 or 2 or 3) = 1/9 + 2/9 + 1/9 = 4/9 = 0.4444.
c. Now, P(O1) + P(O2) + P(O3) + P(O4) + P(O5) + P(O6) = 1 implies that

c + 2c + 3c + 4c + 5c + 6c = 1, from which we get c = 1/21.
Hence P(O1) = 1/21; P(O2) = 2/21; P(O3) = 3/21; P(O4) = 4/21; P(O5) = 5/21; P(O6) = 6/21.
P(odd number) = 1/21 + 3/21 + 5/21 = 9/21 = 0.428571.
P(at most 3) = 1/21 + 2/21 + 3/21 = 6/21 = 0.285714.
P (E ∩ F ) 0.54
6.28 a. P (E | F ) = = = 0.90.
P (F ) 0.6
P (E ∩ F ) 0.54
b. P (F | E ) = = = 0.7714.
P (E ) 0.7
6.29 a. Let E = the event that critic 1 gave a “thumbs up” to the chosen movie and F = the event that
critic 2 gave a “thumbs up” to the selected movie. Then P(E ) = 15/60 = ¼, P(F ) = 20/60 =
1/3, and P( E ∩ F ) = 10 / 60 = 1 / 6.
P (E ∩ F ) 1/ 6
So, the required probability = P (F | E ) = = = 0.6667.
P (E ) 1/ 4
b. Using the given information we construct the following table showing the various possible
outcomes and their frequencies of occurrence.
Critic 2 Critic 2 Total

“thumbs up” “thumbs down”
Critic 1 “thumbs up” 10 5 15
Critic 1 “thumbs down” 10 35 45
Total 20 40 60
35
P (thumbs down from critic 1|thumbs down from critic 2) = = 0.875.
40
6.30 Calculating the totals gives: Total adults aged 18-29: 50

Total adults aged 30-50: 50
Total adults with tattoo: 24
Total without tattoo: 76
Total in survey: 100
24
a. P (tattoo ) = = .24
100
18
b. P (tattoo | age 18 − 29) = = .36
50
6
c. P (tattoo | age 30 − 50) = = .12
50
18
d. P (age 18 − 15 | tattoo ) = = .75
24
136
6.31 The statement in the article implies the following two conditions:
(1) P (D | Y c ) > P (D | Y ) and
(2) P (Y ) > P (Y c ) .
This claim is consistent with the information given in I but not with any of the others. In II and III,
condition (1) is violated; in IV, condition (2) is violated; in V and VI, both conditions are violated.
6.32 It appears to be true that most basketball players are over 6 feet tall, but many people over 6 feet tall
are not basketball players. So it is reasonable to expect P ( A | B ) > P ( B | A) .
432
6.33 a. P(F|FP)= = 0.769
562
390
b. P(M|MP)= = 0.890
438
c. Ultrasound appears to be more reliable in predicting boys
6.34 1. P(M) = 0.021 2. P(M|B) = 0.043 3. P(M|W) = 0.07
135
6.35 a. i. Required conditional probability = = 0.74176.
182
173 + 206
ii. Required conditional probability = = 0.8594.
210 + 231
b.
Properly restrained
Iowa City (city)
Not Properly restrained
Properly restrained
Iowa City (interstate)
Not properly restrained
6.36 The column total for “very harmful” is 224 and the row totals for “current smoker” is 96, “former
smoker” is 99 and “never smoked” is 99. There were 294 people who questioned.
99 224 60 78 86
a. = 0.3367 b. = 0.7619 c. = 0.625 d. = 0.7879 e. = 0.8687
294 294 96 99 99
f . The probabilities in c,d and e differ in that the probability that a randomly selected individual from
that group perceives smoking as very harmful decreases if he or she is a former smoker and
decreases even more if he or she has never smoked. This is probably not a surprise.
6.37 a. 0.1 + 0.175 = 0.275

0.1
b. = 0.2
0.5
0.325
c. = 0.65
0.5
137
0.325
d. = 0.448
0.725
e. The probabilities in c and d are not equal because the events “wearing a seatbelt” and “being
female” are not independent. It is clearly more likely an adult will regularly use a seatbelt if
they are female than if they are male.
100 41 41 500
6.38 a. i. =0.167 ii. = 0.068 iii. = 0.241 iv. = 0.833
600 600 172 600
41 18
v. = 0.41 vi. = 0.18
100 100
b. This clearly shows that the probability of randomly chosen senior citizen using a seatbelt is
far lower than someone in the 18 to 24 year age group.
456 215 241

6.39 a. i. P(S) = = 0.76 ii. P(S|A) = = 0.717 iii. P(S|B) = = 0.803
600 300 300
iv. Drug B appears to have a higher survival rate
140 120 20
b. i. P(S) = = 0.583 ii. P(S|A) = = 0.600 iii. P(S|B) = = 0.5
240 200 40
iv. Drug A appears to have the higher survival rate.
316 95 221
c. i. P(S) = = 0.878 ii. P(S|A) = = 0.95 iii. P(S|B) = = 0.85
360 100 260
iv. Drug A appears to have the higher survival rate
d. Women seem to respond well to both treatments. Men seem to respond very well to
treatment A but not so well to treatment B. When the data is combined, the small quantity of
data collected for men on treatment B - only 20% of those men who had treatment A,
becomes “lost” in the rest of the data.
6.40 Suppose the probability that a driver that does not live close to a restaurant will have an accident is
x . So P( A | RC ) = x . If a driver who does live close to a restaurant is 30% more likely to have an
accident, then P ( A | R ) = 1.3 x . The difference between these is 0.3x which divided by x; P( A | R ) ,
C
will give 0.3. The 4th choice is the correct one.
6.41 a. About 85% of all the past calls were for medical assistance.
b. P(call is not for medical assistance) = 1 – 0.85 = 0.15.
c. P(two successive calls are both for medical assistance) = (0.85 )(0.85) = 0.7225.
d. P(first call is for medical assistance and second call is not for medical assistance)
= (0.85)(0.15) = 0.1275.
e. P(exactly one of two calls is for medical assistance) = P(first call is for medical assistance
and the second is not) + P(first call is not for medical assistance but the second is) =
(0.85)(0.15) + (0.15)(0.85) = 0.255.
f. Probably not. There are likely to be several calls related to the same event—several reports
of the same accident or fire that would be received close together in time.
138
6.42 a. In the long run, 30 out of every 100 accidents reported to the police that result in a death
were a single-vehicle rollover.
b. In the long run, 54 out of every 100 accidents reported to the police that result in a death
were a frontal collision.
c. P(R|D) = 0.3, P(R) = 0.06. As P(R|D)≠P(R), the events R and D are not independent.
d. P(F∩D) is only equal to P(F)P(D) if the events F and D are independent. As P(F|D)≠P(F):
0.54≠0.6 , the events F and D are not independent⇒ P(F∩D) ≠ P(F)P(D).
e. No, F is the event that the selected accident was a frontal collision. Rc is the event that the
selected accident was not a single-vehicle rollover. These are not the same. There are
accidents that are one of these, or both or neither.
6.43 No, a randomly selected adult is more likely to experience pain daily if they are a woman. The events
are dependent.
6.44 Let A = the event that the selected student (from all kindergartners who were tested for TB) has TB;
B = the event that the selected student (selected from all kindergartners who were tested for TB) is a
recent immigrant. Based on the information given, we have P(A | B) = 0.0075 and P(A) = 0.0006. For
independence, we need P(A | B) = P(A) which is not true here. So the two outcomes under
consideration are dependent.
6.45 Let A = the event that the selected adult is a female and B = the event that the selected adult favors
stricter gun control. Based on the given information we estimate P(B | A) = 0.66 and P(B) = 0.56.
Since P(B | A) and P(B) are not equal, we conclude that the two events A and B are dependent
outcomes.
6.46 Let A = the event that the selected smoker who is trying to quit uses a nicotine aid and B = the event
that the selected smoker who has attempted to quit begins smoking again within two weeks. From
the given information we can deduce the following.
P ( A) = 0.113 ; P( B | A c ) = 0.62 ; P ( B | A) = 0.6 . So P ( A ∩ B ) = P ( A) P ( B | A) = 0.0678.
From this we get,

P( A c ∩ B) = P( A c ) P( B | A c ) = (1 − 0.113)(0.62) = 0.54994.
Therefore
P( B) = P( A ∩ B) + P( A c ∩ B) = 0.61774.
Since P( B | A) is not equal to P (B ) , A and B are dependent outcomes.
6.47 P(jury duty call two years in a row) = (0.15)(0.15) = 0.0225.

P(jury duty call three years in a row) = (0.15)(0.15)(0.15) = 0.003375.
6.48 a. Assuming that whether Jeanie forgets to do one of her ‘to do’ list items is independent of
whether or not she forgets any other of her ‘to do’ list items, the probability that she forgets all
three errands = (0.1)(0.1)(0.1) = 0.001.
b. P(remembers at least one of her three errands) = 1 – P(she forgets all three errands)
= 1 – 0.001 = 0.999.
c. P(remembers the first errand, but not the second or the third) = (0.9)(0.1)(0.1) = 0.009.
139
6.49 a. P(none of the 10 calls lead to a reservation) = (0.7)10 = 0.0282475249.
b. It was assumed that the outcomes of the different calls are independent, i.e., whether one
call
resulted in a reservation or not has no influence on the results of other calls.
c. P(at least one call leads to a reservation) = 1 – 0.0282475249 = 0.9717524751.
6.50 a. P(student has a Visa card) = 7000/10000 = 0.7
b. P(student has both cards) = 5000/10000 = 0.5
c. P(student has both cards | student has Visa card) = 5000/7000 = 0.714286
d. P(student has a Visa card & a Master Card) = 0.5. For the two events to be independent we
must have P(student has a Visa card & a Master Card ) = P(has Visa card) P(has a Master
Card) = (7000/10000) (6000/10000) = 0.42 which is not equal to 0.5. So the two events are
dependent.
e. If only 4200 has both cards, then the condition P(student has a Visa card & a Master Card ) =
P(has Visa card) P(has a Master Card) holds, so the two events are independent.
6.51 a. P( 1-2 subsystem works) = P(1 works) P(2 works) = (0.9)(0.9) = 0.81.
b. P( 1-2 subsystem does not work ) = 1 – P( 1-2 subsystem works) = 1 – 0.81 = 0.19.
P( 3-4 subsystem does not work) = 1 – P( 3-4 subsystem works) = 1 – (0.9)(0.9) = 0.19.
c. Using the results of part b, P( system won’t work) = P( 1-2 subsystem doesn’t work & 3-4
subsystem doesn’t work ) = P( 1-2 subsystem doesn’t work ) P( 3-4 subsystem doesn’t work)
= (0.19)(0.19) = 0.0361.
So P( system will work ) = 1 – P( system won’t work) = 1 – 0.0361 = 0.9639.
d. If a 5-6 subsystem were added, then P( system will work ) = 1 – P( system won’t work )
= 1 – (0.19)(0.19)(0.19) = 0.993141.
e. If there were 3 components in series in each subsystem, then P( system will work )
= 1 – P( system won’t work) = 1 – P( first subsystem doesn’t work) P( second subsystem
doesn’t work). The probability that the first subsystem doesn’t work = 1- (0.9)(0.9)(0.9) = 1 –
0.729 = 0.271. This is also the probability that the second subsystem won’t work. So
P(system works) = 1 – (0.271)(0.271) = 0.926559.
6.52 a. i. P(F) = 0.51

ii. P(F | C) = 0.56
iii. P(F | Cc ) = 0.45
iv. P(F | O) = 0.36
v. P(F | Y) = 0.72
b. P(F|C) is not equal to P(F) so F and C are dependent.
c. P(F|O) is not equal to P(F) so F and O are dependent.
6.53 a. The expert assumed that a valve stem could be in any one of the “clock positions” with equal
probability, so that there is a 1 in 12 chance for each valve to be in any given clock hour
position. In addition, the expert assumed that the positions of the two valve stems are
independent. Hence, the expert computed the probability of a “match” on both valves to be
the product (1/12)(1/12) = 1/144.
140
b. The assumption of “independence” of the two valve stems is highly questionable because the
valve stems turn in a “synchronized” manner. This would make the probability of the valve
stems being in their original positions greater than the 1/144 value computed by the expert.
6.54 a. P(A wins both matches and B beats C) = P(A beats B) P(A beats C) P(B beats C) =
(0.7)(0.8)(0.6) = 0.336.
b. P(A wins both her matches) = P(A beats B) P(A beats C) = (0.7)(0.8) = 0.56.
c. P(A loses both her matches) = P(A loses to B) P(A loses to C) = (1 – 0.7)(1 – 0.8) = 0.06.
d. There are two ways this can happen – A beats B, B beats C, C beats A OR A beats C, B
beats A, and C beats B. So the required probability = (0.7) (0.6) (1 – 0.8) + (0.8) ( 1 – 0.7) (1
– 0.6) = 0.084 + 0.096 = 0.18.
6.55 a. If the first board is defective, then the probability that the second board is defective =
39/4999. If the first board is not defective, then the probability that the second board is
defective = 40/4999. Since the probability that the second board is defective changes
depending on whether or not the first board is defective, the events E1 and E2 are dependent.
b. P(not E1 ) = 1 – P(E1) = 1 - (40/5000) = 0.992.
c. P(E2 | E1 ) = 39/4999 and P(E2 | not E1 ) = 40/4999. They are nearly equal.
d. Yes.
6.56 P(small size of brand B1 ) = P(small) P(brand B1 ) = (0.30)(0.40) = 0.12. The other probabilities are
obtained in a similar manner. The results are shown in the following table.
Size
S M L
B1 0.12 0.20 0.08 0.40
Brand
B2 0.18 0.30 0.12 0.60
0.30 0.50 0.20 1.00
6.57 a. P(CC) = (50/800)(50/800) = 0.00391.
b. P(CC) = P(first answer is correct) P(second answer is correct | first answer is correct)
= (50/800)(49/799) = 0.00383. This is very close to the probability computed in part a.
6.58 For sampling with replacement, P(both answers are correct) = (50/100)(50/100) = 0.25.
For sampling without replacement P(both answers are correct) = (50/100) (49/99) = 0.24747.
Again, the two probabilities are quite close to each other, but not as close as the two probabilities in
parts b and c of Problem 6.57..
6.59 a. P(E) = 6/10 = 0.6
b. P(F | E) = 5/9 = 0.5556
c. P(E and F) = P(E) P(F|E) = (6/10)(5/9) = 30/90 = 1/3
141
6.60 a. P(E1 and E2) = (0.4)(0.3) = 0.12
b. P(not E1 and not E2) = [1 – P(E1)][1 - P(E2)] = (0.6)(0.7) = 0.42
c. P(successful in at least one of the two bids) = 1 – (both bids unsuccessful)

= 1 – 0.42 = 0.58. (using the result of part b)
6.61 a. P(individual has to stop at at least one light) = P(E ∪ F) = P(E) + P(F) – P(E ∩ F)
= 0.4 + 0.3 – 0.15 = 0.55.
b. P(individual doesn’t have to stop at either traffic light) = 1 – P(must stop at at least one light)
= 1 – 0.55 = 0.45 (using the result of part a).
c. P(stop at exactly one of the two lights) = P(stop at least one of the two lights) – P(stop at both
lights) = 0.55 – 0.15 = 0.40.
d. P(E only) = P(E and not F) = P(E) – P(E ∩ F) = 0.4 – 0.15 = 0.25.
6.62 The event E ∩ F means that a randomly selected registered voter in a certain city has signed a
petition to recall the mayor and also actually votes in the recall election. P(E ∩ F) = P(F) P(E | F) =
(0.10)(0.80) = 0.08.
6.63 The likelihood of using a cell phone is different for of each type of vehicle; for instance if you drive a
van or SUV, you are more likely to use a cell phone than if you drive a pick-up truck.
4.5 4.1
6.64 a. P(F) = = 0.523, P(S) = = 0.477, P(R|F) = 0.018, P(R|S) = 0.03
8.6 8.6
P (S )P (R | S ) 0.01431
b. P(S|R) = = = 0.6032
P (S )P (R | S ) + P (F )P (F | S ) 0.01431 + 0.0094
This is the probability that a randomly selected gun purchase background check resulted in a
blocked sale by a state or local agency.
6.65 a. i. 0.5 ii. 0.5 iii. 0.99 iv. .01
b. P(TD) = P(TD∩D) + P(TD∩C) = P(TD|D)P(D) + P(TD|C)P(C) =

= (.99)(.01) + (.99)(.01) = 2(.99)(.01) = .0198
P (C )P (TD | P (C ) (.99)(.01)
c. P(C|TD) = = = 0.5
P (TD | P (C ) + P (TD ) | P (D ) 2(.99)(.01)
Yes, it is consistent with the argument given in the quote.
6.66 Letting P( E ∩ C ∩ L) = x , P ( E ∪ C ∪ L) = .49 so using the other segments of a Venn Diagram and
solving for x :
(.40-.23-.22+ x ) + (.23- x ) +(.30-.23-.21+ x ) +(.22- x ) + x +(.21- x )+(.25-.22-.21+ x ) =.49
x =20.
a P ( E ) = .4 b P (C ) = .3 c P( L) = .25 d P ( E and C) =.23
e P( E c and C c and Lc ) = .51 f P ( E or C or L ) = .49 g P ( E | L) = .88
h P ( L | C ) = .70 i P ( E and C and L) = .20 j P ( E and L ) = .22
k P (C and L ) = .21 l P (C | E and L ) = .91
142
6.67 a. In 4 consecutive years, there are 3 years of 365 days (1095 non leap days) and 1 year of 366
days (365 non leap days and one leap day) – a total of 1461 days, of which one, Feb 29th is a
leap day. Hence a leap day occurs once in 1461 days.
b. If babies are induced or born by C-section, they are less likely to be born at weekends or on
holidays so they are not equally likely to be born on the 1461 days.
c. 1 in 2.1 million is the probability of picking a mother and baby randomly and finding both to be
a leap-day mom-baby (1/14612). The probability of a leap year baby becoming a leap year
mom means that the mom’s birthday is already fixed, so the hospital spokesperson’s
probability is too small.
6.68 a. Let E = the event that the selected employee uses drugs and F = the event that the selected
employee tests positive. Then P(F|Ec ) = P(false positive) = 0.05 and P(Fc |E) = 0.10. So
P(F|E) = 1 – P(Fc |E) = 1 – 0.10 = 0.90. Also, P(E) = 0.10. We need P(E and F). We have P(E
and F) = P(E)P(F | E) = (0.10)(0.90) = 0.09.
b. P(Ec ∩ F) = P(Ec)P(F | Ec ) = (1 – 0.10)(0.05) = (0.9)(0.05) = 0.045.
c. P(F) = P(E ∩ F) + P(Ec ∩ F). Now, P(E ∩ F) = 0.09 from part a and P(Ec ∩ F) = 0.045
from part b. Therefore P(F) = 0.09 + 0.045 = 0.135.
P(E ∩ F) 0.09
d. P(uses drugs | tests positive) = P(E|F) = = = 0.67 , using the results from
P(F) 0.135
parts a and c.
6.69 a. Assume that the results of successive tests on the same individual are independent of one
another. Let F1 = the event that the selected employee tests positive on the first test and F2 =
the event that he/she tests positive on the second test. Then, P(employee uses drugs and
test positive twice) = P(E)P(F1|E)P(F2|E) = (0.1)(0.90)(0.90) = 0.081.
b. P(employee tests positive twice)

= P(employee uses drugs and tests positive twice)
+ P(employee doesn’t use drugs and tests positive twice)
= P(E) P(F1|E)P(F2|E) + P(Ec) P(F1|Ec)P(F2|Ec)
= (0.1)(0.90)(0.90) + (0.9)(0.05)(0.05) = 0.081 + 0.00225 = 0.08325.
c. P(uses drugs | tested positive twice) =

P(uses drugsandtestspositive twice) 0.081
= = 0.97297.
P(testspositive twice) 0.08325
d. P(tests negative on the first test OR tests positive on the first test but tests negative on the
second test | does use drugs) = P(F1c | E ) + P(F1 and F2c |E )
= [1 – P(F1 |E )] + P(F1 |E ) [1 – P(F2 |E )]
= [1 – 0.9] + (0.9)(1 – 0.9) = 0.1 + 0.09 = 0.19
(using P(F1 |E ) = 0.90, from part a of Problem 6.68).
e. The benefit of using a retest is that it reduces the rate of false positives from 0.05 to 0.02703
(obtained as 1 – 0.97297, using the result from part c), but the disadvantage is that the rate
of false negatives has increased from P(Fc |E) = 0.10 to 0.19 (see part d above). Retests also
involve additional expenses.
143
6.70 a. Let a = the number of adults who are full-time workers and drug users, b = number of adults
who work full-time and do not use drugs, c = number of adults who do not work full-time but
use drugs, and d = number of adults who do not work full time and do not use drugs. The
given statements imply that a/(a+b) = 0.08 and a/(a+c) = 0.7. This is possible if
b = (0.92/0.08)a = (11.5)a and a = (7/3)c. So, this is indeed possible – for instance, take
c = 30, a = 70, b = 805, d = 7095. See the following table for an example. (There are other
possible values of a, b, c, d, for which the given statements hold also.)
Works full-time Does not work full-time Total

Use drugs 70 30 100
Does not use drugs 805 7095 7900
Total 875 7125 8000
b. Based on the given information, P(D|E) is estimated to be 0.08 and P(E|D) is estimated to be
0.70.
c. It is not possible to determine P(D) from the information given because the given information
translates into the two equations -- a/(a+b) = 0.08 and a/(a+c) = 0.7, which do not have a
unique solution. See part a. If, for instance, any one of the values a, b, c, d is given, then we
can determine the rest of the values and thereby obtain P(D) = (a+b)/(a+b+c+d) =
(a+b)/8000.
6.71 Based on the given information we can construct the following table.
Basic Deluxe Total
Model Model
Buy extended warranty 12% 30% 42%
Do not buy extended warranty 28% 30% 58%
Total 40% 60% 100%
So P(Basic model | extended warranty) = 12/42 = 0.285714.
6.72 a. P(satisfied) = P(student took the class during fall quarter and was satisfied with book) +
P(student took the class during winter quarter and was satisfied with the book) + P(student
took the class during spring quarter and was satisfied with the book) = (200/1000) +
(150/1000) + (160/1000) = 0.51.
b. P(took the class during fall quarter | satisfied with book)
P(took the class during fall & satisfied) (200 /1000) 0.2
= = = = 0.3922 .
P(satisfied) 0.51 0.51
P(took the class during winter quarter | satisfied with book)
P(took the class during winter & satisfied) (150 /1000) 0.15
= = = = 0.2941 .
P(took the class during spring quarter | satisfied with book)
P(took the class during spring & satisfied) (160 /1000) 0.16
= = = = 0.3137 .
Hence, given that the selected student was satisfied with the textbook, the most likely book
used was by Professor Mean. This has the highest conditional probability (0.3922) among the
three possibilities. The least likely would be the book by Professor Median.
144
6.73 Let A = the event that he takes the small car; B = the event that he takes the big car; C = the event
that he is on time. Then, based on the given information we have P(A) = 0.75, P(B) = 0.25, P(C|A) =
0.9, and P(C|B) = 0.6. We need P(A|C). Using Bayes’ Rule we get
P(C|A)P(A) (0.9)(0.75)
P(A|C) = = = 0.8182.
P(C|A)P(A) + P(C|B)P(B) (0.9)(0.75) + (0.6)(0.25)
6.74 a. Let A = the event that the selected individual has the disease and B = the event that the test
result is positive. Then it is given that P(A) = 0.001, P(B|A) = 0.95, and P(Bc|Ac) = 0.90.
So P(B|Ac)= 1 – 0.90 = 0.10.
Positive Test
0.95
Has disease
0.001
0.05
Negative Test
Positive Test
0.10
0.999
Doesn’t have disease
0.90
Negative Test
b. P( has disease and positive test) = (0.001)(0.95) = 0.00095.
c. P(positive test) = P(has disease and positive test) + P(doesn’t have disease and positive test)
=(0.001)(0.95) + (0.999)(0.10) = 0.00095 + 0.0999 = 0.10085.
d. Using the definition of conditional probability, we get

P(has disease & positive test) 0.00095
P(has disease | positive test) = = = 0.00942.
P(positive test) 0.10085
Yes, the result is surprising because, the test has a fairly high probability of coming up
positive when the individual has the disease and a fairly small probability of coming up
positive when the individual doesn’t have the disease. So, given that the test result is
positive, one would have thought there is a fairly high chance that the subject has the
disease. However, this probability is only 0.00942. The reason this probability is so small is
that the incidence rate of the disease is very small and the positive test result is getting
“downweighted” because of this prior knowledge that the disease is “rare”.
6.75 a. P(on time delivery in Los Angeles) = 425/500 = 0.85.
b. P(late delivery in Washington, D.C) = (500 – 405)/500 = 95/500 = 0.19.
c. Assuming that whether or not one letter is delivered late is “independent” of the on-time
delivery status of any other letter, we calculate P(both letters delivered on-time in New York)
= (415/500)(415/500) = 0.6889. (The independence assumption may not be a valid
assumption).
d. P(on-time delivery nationwide) = 5220/6000 = 0.87.

145
6.76 a.
High School GPA
Probatio 2.5 -< 3.0 -< 3.5 3.5 and Above Total
n 3.0
Yes 0.10 0.11 0.06 0.27
No 0.09 0.27 0.37 0.73
Total 0.19 0.38 0.43 1.00
b. P(academic probation) = 0.10 + 0.11 + 0.06 = 0.27.
c. P(high GPA of 3.5 or above) = 0.43.
d. P(GPA of 3.5 or above AND academic probation) = 0.06.

P(GPA of 3.5 or above)P(academic probation) = (0.43)(0.27) = 0.1161.
Since P(GPA of 3.5 or above AND academic probation) is not equal to P(GPA of 3.5 or
above)P(academic probation), we conclude that the events are dependent.
e. P(GPA between 2.5 and 3.0 | academic probation) = 0.10/0.19 = 0.5263
f. P(GPA of 3.5 or above AND academic probation) = 0.06/0.43 = 0.1395.
6.77 a. P(male) = (200+800+1500+1500+900+1500+300)/17000 = 6700/17000 = 0.3941.
b. P(agriculture) = 3000/17000 = 0.1765.
c. P(male AND from agriculture) = 900/17000 = 0.0529.
d. P(male AND not from agriculture) = 5800/17000 = 0.3412.
6.78 Total Monterey County = 369,000 Total Caucasian = 1,003,000

Total San Luis Obispo County = 236,000 Total Hispanic = 528,000
Total Santa Barbara County = 392,000 Total Black = 61,000
Total Ventura County = 736,000 Total Asian = 122,000
Total American Indian = 19,000
Total Count = 1,733,000
736,000
a. = 0.4247
1,733,000
231,000
b. = 0.3139
736,000
231,000
c. = 0.4375
528,000
9000
d. = 0.0052
1,733,000
122000 236000 9000

e. P (Asian or from San Luis Obispo) = + − = 0.2014
1,733,000 1,733,000 1,733,000
146
f. P (Asian or from San Luis Obispo, but not both) =
122000 236000 18000
+ − = 0.1962
1,733,000 1,733,000 1,733,000
1,003,000 1,002,999
g. P (both Caucasian) = ( )( ) = 0.335
1,733,000 1,732,999
730000 729999
h. P (neither Caucasian) = ( )( ) = 0.1774
1733000 1732999
i. P (exactly one Caucasian) = P (first is Caucasian, second is not) + P (first is not

Caucasian, second is Caucasian) =
1003,000 730000 730000 1003000
( )( )+( )( ) = .2438 + .2438 = 0.4876
1733000 1732999 1733000 1732999
j. P (both from same county) = P (both from Monterey) + P (both from San Luis Obispo) +
P ( both from Santa Barbara) + P (both from Ventura) =
369,000 368,999 236,000 235,999 392000 391999 736000 735999
( )( )+( )( )+( )( )+( )( )
1,733,000 1,732,999 1,733,000 1732999 1733000 1732999 1733000 1732999
= 0.0453 + 0.0185 + 0.0512 + 0.1804 = 0.2954
k. P (both from different racial/ethnic group) = 1 – P (both from the same racial/ethnic
group = 1 – P (both Caucasian) – P (both Hispanic) – P (both Black) – P (both Asian) – P
(both American Indian)
528,000 527999
P (both Caucasian) = 0.335 P (both Hispanic) = ( )( ) = 0.0928
1733000 1732999
61000 60999
P (both Black) = ( )( ) = 0.0012
1733000 1732999
122000 121999
P (both Asian) = ( )( ) = 0.0050
1733000 1732999
19000 18999
P (both American Indian) = ( )( ) = 0.0001
1733000 1732999
P (both from different racial/ethnic groups) =

1- 0.335 – 0.0928 – 0.0012 – 0.0050 – 0.0001 = 0.5659
6.79 The results will vary from one simulation to another. The approximate probabilities given were
obtained from a simulation done by computer with 100,000 trials.
a. 0.71293 b. 0.02201
6.80 a. Assume that the requests are numbered from 1 to 20 as follows:

1−5 requests for a single license
6−14 requests for 2 licenses
15−20 requests for 3 licenses
Select a two digit random number from the table. If it is over 20 ignore it. If it is a number
from 1 to 20 retain it. If it is a number from 1 to 5 then 1 license will be granted on this
selection. If it is a number from 6−14 then two licenses will be granted for this selected
number. If it is a number from 15−20 then three licenses will be granted on this selected
number.
Select a second random number between 1 and 20 and grant the license request.
Select a third random number between 1 and 20 and grant the license request.
147
Select a fourth random number between 1 and 20 and grant the request if there are enough
licenses remaining to satisfy the request.
Repeat the previous step until all 10 licenses have been granted.
The simulation results will vary from one simulation to another. The approximate probability
obtained from a simulation with 20,000 trials done by computer was .3467.
b. An alternative procedure for distributing the licenses might be as follows. The twenty
individuals requested a total of 41 licenses. Create a box containing 41 slips of paper. If an
individual requested 1 license, his name is placed on only 1 piece of paper. If an individual
requested two licenses, his name is placed on two slips. If he requested three licenses, his
name is placed on three slips. Choose 10 slips of paper at random from the collection.
Licenses are granted to those individuals whose names are selected. A person is granted as
many licenses as the number of times their name is selected.
6.81 The simulation results will vary from one simulation to another. The approximate probability should
be around .8504.
6.82 The simulation results will vary from one simulation to another. The approximate probability should
be around .8468.
6.83 a. The simulation results will vary from one simulation to another. The approximate probability
should be around 0.6504.
b. The decrease in the probability of on time completion for Jacob made the biggest change in
the probability that the project is completed on time.

R
6.84 a.
0.05
Line A1
0.95
0.5 N
R
0.08
0.3
Line A2
0.92
N
0.2
0.10 R
Line A3
0.90
148
b. P(came from line 1 and needs rework) = (0.5)(0.05) = 0.025.
c. P(needs rework) = P(came from line 1 and needs rework) + P(came from line 2 and needs
rework) + P(came from line 3 and needs rework) = (0.5)(0.05) + (0.3)(0.08) + (0.2)(0.1) =
0.069.
6.85 P(parcel went via A1 and was late) = P(went via A1)P(late | went via A1) = (0.4)(0.02) = 0.008.
6.86 a. P(arrived late) = P(sent via A1 and arrived late) + P(sent via A2 and arrived late) + P(sent via
A3 and arrived late) = (0.4)(0.02) + (0.5)(0.01) + (0.10)(0.05) = 0.018.
b. We use the definition of conditional probability to compute each of the following.
P(sent via A1 and arrived late) (0.4)(0.02) 4

P(A1 |L) = = = = 0.44444 .
P(arrived late) 0.018 9
P(sent via A 2 and arrived late) (0.5)(0.01) 5

P(A2 |L) = = = = 0.27778 .
P(sent via A 3 and arrived late) (0.1)(0.5) 5

P(A3 |L) = = = = 0.27778 .
6.87 a. (.3 )5 = .00243
b. (.3 )5 + (.2)5 = .00243 + .00032 = .00275

c. Select a random digit. If it is a 0, 1, or 2, then award A one point. If it is a 3 or 4, then award
B one point. If it is 5, 6, 7, 8, or 9, then award A and B one-half point each. Repeat the
selection process until a winner is determined or the championship ends in a draw (5 points
each). Replicate this entire procedure a large number of times. The estimate of the
probability that A
wins the championship would be the ratio of the number of times A wins to the total number
of replications.
d. It would take longer if no points for a draw are awarded. It is possible that a number of
games would end in a draw and so more games would have to be played in order for a player
to earn 5 points.
6.88 a. Choose a random digit. If it is a 1, 2, 3, 4, 5, 6, 7 or 8, then seed 1 defeats seed 4. If it is a 9

or 0, then seed 4 defeats seed 1.
b. Choose a random digit. If it is a 1, 2, 3, 4, 5 or 6, then seed 2 defeats seed 3. If it is a 7, 8, 9,

or 0, then seed 3 defeats seed 2.
c. If seeds 1 and 2 have won in the first round, then they are competing in game 3. Select a
random digit. If it is a 1, 2, 3, 4, 5 or 6, then seed 1 defeats seed 2 in game 3. If it is a 7, 8,
9, or 0, then seed 2 defeats seed 1 in game 3.
If seeds 1 and 3 have won in the first round, then they are competing in game 3. Select a
random digit. If it is a 1, 2, 3, 4, 5, 6, or 7, then seed 1 defeats seed 3 in game 3. If it is an 8,
149
random digit. If it is a 1, 2, 3, 4, 5, 6, or 7, then seed 2 defeats seed 4 in game 3. If it is an 8,
random digit. If it is a 1, 2, 3, 4, 5, or 6, then seed 3 defeats seed 4 in game 3. If it is a 7, 8,
d. For game 1, selected digit is 7. Seed 1 wins game 1.

For game 2, selected digit is 9. Seed 3 wins game 2.
So seeds 1 and 3 compete in game 3.
For game 3, selected digit is 6. Seed 1 wins game 3 and the tournament.
e. Summarize the result of part d as follows: digits selected (7,9,6) → seed 1 wins tournament.
Replication Digits Winner of

s Selected Tournament
2 (5,2,9) seed 2
3 (0,6,0) seed 4
4 (7,5,8) seed 2
5 (0,4,1) seed 2
6 (6,6,9) seed 2
7 (3,1,2) seed 1
8 (2,6,2) seed 1
9 (7,5,0) seed 2
10 (5,6,4) seed 1
Based on this simulation of 10 trials, the estimate of the probability that the first seed wins the
tournament is 4/10 = .4.
f. Answers will vary.
g. The answers for parts e and f differ because they are based on different random selections
and different number of trials. Generally, the larger the number of trials, the better the
estimate is. So one would believe that the estimate from part f is better than the estimate of
part e.
6.89 P(need to examine at least 2 disks) = 1 – P(get a blank disk in the first selection) = 1 – 15/25 = 0.4.
6.90 a. P(Ec) = 1 – 0.60 = 0.40
b. Events E and F are disjoint, so P(E ∪ F) = P(E) + P(F) = 0.60 + 0.15 = 0.75.
c. P(Ec ∩ Fc) = P( (E ∪ F)c ) = 1 – P(E ∪ F) = 1 – 0.75 = 0.25.
6.91 Let the five faculty members be A, B, C, D, E with teaching experience of 3, 6, 7, 10, and 14 years
respectively. The two selected individuals will have a total of at least 15 years of teaching experience
if the selected individuals are A and E, or B and D, or B and E, or C and D, or C and E, or D and E.
Since there are a total of 10 equally likely choices and 6 out of these meet the specified requirement,
P(two selected members will have a total of at least 15 years experience) = 6/10 = 0.6.
6.92 P(reads at least one of the three magazines) = P( A or B or C)

= 0.14 + 0.23 + 0.37 – 0.08 – 0.09 – 0.13 + 0.05 = 0.49.
150
6.93 Total number of viewers = 2517.
179 + 87
a. P(saw a PG movie) = = 0.1057.
2517
420 + 323 + 179 + 114 + 87
b. P(saw a PG or a PG-13 movie) = = 0.4462.
2517
2517 − 600 − 196 − 205 − 139

c. P(did not see an R movie) = = 0.5471.
2517
6.94 R1 and R2 are not independent events because P(R2|R1) = 1139/2516 and this is not equal to
P(R2|R1c) = 1140/2516. However, these two probabilities are very nearly equal, so from a practical
point of view these two events may be regarded as independent (approximately).
6.95 a. P(neither is selected for testing | batch has 2 defectives) = (8/10)(7/9) = 56/90 = 0.6222.
b. P(batch has 2 defectives and neither selected for testing) = P(batch has 2 defectives)
P(neither is selected for testing | batch has 2 defectives) = (0.20)(0.6222) = 0.1244.
c. P(neither component selected is defective) = P(batch has 0 defectives & neither selected
component is defective) + P(batch has 1 defective & neither selected component is defective)
+ P(batch has 2 defectives & neither selected component is defective) = (0.5)(1) +
(0.3)(9/10)(8/9) + (0.2)(8/10)(7/9) = 0.5 + 0.24 + 0.1244 = 0.8644.
6.96 a. P(E) = 20/25 = 0.8.
b. P(F|E) = 19/24 = 0.79167.
c. P(G|E ∩ F) = 18/23 = 0.78261.
d. P(E ∩ F ∩ G) = P(E) P(F | E) P(G | E ∩ F) = (20/25)(19/24)(18/23) = 0.49565.
6.97 P(E ∩ F ∩ G ∩ H) = P(E)P(F|E)P(G| E ∩ F)P(H| E ∩ F ∩ G) = (20/25)(19/24)(18/23)(17/22) = 0.3830.

P(at least one bad bulb ) = 1 – P(all 4 bulbs are good ) = 1 – 0.383 = 0.617.
6.98 a. P(1 is sent by all relays | 1 sent by transmitter) = (0.8)(0.8)(0.8) = 0.512.
b. P(1 is received by the receiver | 1 is sent by transmitter) = P(relay 1 sends 1, relay 2 sends 1,
relay 3 sends 1) + P(relay 1 sends 1, relay 2 sends 0, relay 3 sends 1) + P(relay 1 sends 0,
relay 2 sends 1, relay 3 sends 1) + P( relay 1sends 0, relay 2 sends 0, relay 3 sends 1) =
(0.8)(0.8)(0.8) + (0.8)(0.2)(0.2) + (0.2) (0.2) (0.8) + (0.2)(0.8)(0.2) = 0.608.
151
Chapter 7
7.1 a. discrete
b. continuous
c. discrete
d. discrete
e. continuous
7.2 a. continuous
b. continuous
c. continuous
d. discrete
e. continuous
f. continuous
g. discrete
7.3 The possible y values are the set of all positive integers.
Some possible outcomes are LS, RRS, S, LRRRLLRLLS, and LLLLS, with corresponding y values
equal to 2, 3, 1, 10, and 5, respectively.
7.4 Possible values of x (in feet) are the real numbers in the interval 0 ≤ x ≤ 2 . The variable x is a
continuous variable.
7.5 Possible values of y (in feet) are the real numbers in the interval 0 ≤ y ≤ 100. The variable y is a
continuous variable.
7.6 Possible value of y are the collection of positive even integers 2, 4, 6, 8, etc. The variable y is a
discrete variable.
7.7 a. Possible values for x are 3, 4, 5, 6, 7.
b. If y = first number – second number, then possible values of y are –3, – 2, – 1, 1, 2, and 3.
c. Possible values of z are 0, 1, 2.
d. Possible values of w are 0, 1.
7.8 a. P(x = 4) = 0.25
b. P(x ≤ 4) = 0.02 + 0.03 + 0.09 + 0.25 = 0.39
153
c. P(at most 5 courses) = P(x ≤ 5) = 1 – P(x = 6 or x = 7) = 1 – 0.16 – 0.05 = 0.79
d. P(at least 5 courses) = P(x = 5 or x = 6 or x = 7) = 0.40 + 0.16 + 0.05 = 0.61
P(more than 5 courses) = P(x = 6 or x = 7) = 0.16 + 0.05 = 0.21
e. P(3 ≤ x ≤ 6) = P(x = 3 or x = 4 or x = 5 or x = 6) = 0.09 + 0.25 + 0.40 + 0.16 = 0.90.

P(3 < x < 6) = P(x = 4 or x = 5) = 0.25 + 0.40 = 0.65.
The two probabilities are different because the first probability is for the event that the student
is taking 3 courses, 4 courses, 5 courses, or 6 courses and the second probability is for the
event that the student is taking 4 courses or 5 courses.
7.9 a. p(4) = 1 – p(0) – p(1) – p(2) – p(3) = 1 – 0.65 – 0.20 – 0.10 – 0.04 = 0.01.
b. In the long run, the proportion of cartons that have exactly one broken egg will equal 0.20.
c. P(y ≤ 2) = 0.65 + 0.20 + 0.10 = 0.95. This means that, in the long run, the proportion of
cartons that have two or fewer broken eggs will equal 0.95.
d. P(y < 2) = 0.65 + 0.20 = 0.85. This probability is less than the probability in part c because
the event y = 2 is now not included.
e. P(exactly 10 unbroken eggs) = P(exactly 2 broken eggs) = P(y = 2) = 0.10.
f. P(at least 10 unbroken eggs) = P(0 or 1 or 2 broken eggs) = P(y ≤ 2) = 0.95 (from part c).
7.10 a. Possible outcomes are (1, 2), (1, 3), (1, 4), (2, 3), (2, 4), (3, 4).
b. Each outcome has probability 1/6.
c. The following table gives the value of x for each possible outcome along with the
corresponding probabilities.
Outcome Value of x Probability
(1, 2) 0 1/6
(1, 3) 1 1/6
(1, 4) 1 1/6
(2, 3) 1 1/6
(2, 4) 1 1/6
(3, 4) 2 1/6
From this table we get probability distribution of x which is displayed below.

Value of x Probability
0 1/6
1 4/6
2 1/6
7.11 a. P(airline can accommodate everyone who shows up)

= P(x ≤ 100) = 0.05 + 0.10 + 0.12 + 0.14 + 0.24 + 0.17 = 0.82.
b. P(not all passengers can be accommodated) = P(x > 100) = 1 – P(x ≤ 100) = 1 – 0.82 =
0.18.
c. P(number 1 standby will be able to take the flight) = P(x ≤ 99)

= 0.05 + 0.10 + 0.12 + 0.14 + 0.24 = 0.65.
P(number 3 standby will be able to take the flight) = P(x ≤ 97) = 0.05 + 0.10 + 0.12 = 0.27.
154
7.12 a. The ten different possible outcomes are
(1, 2), (1, 3), (1, 4), (1, 5), (2, 3), (2, 4), (2, 5), (3, 4), (3, 5), (4, 5).
b. The possible values of x are 0, 1, 2. The following table gives the probability distribution of x.
Value of x Probability Boards Selected

0 3/10 (3,4), (3,5), (4,5)
1 6/10 (1,3), (1,4), (1,5), (2,3), (2,4),
(2,5)
2 1/10 (1,2)
7.13 Results will vary. One particular set of 50 simulations gave the following results.
Value of x Frequency Relative Frequency

0 13 13/50 = 0.26
1 32 32/50 = 0.64
2 5 5/50 = 0.1
These relative frequencies are quite close to the theoretical probabilities given in part b of Problem
7.12.
7.14 The possible values of x are 0, 1, 2, 3.
P(x = 0) = P(I, I, I) = (0.3)(0.3)(0.3) = 0.027.

P(x = 1) = P(D, I, I) + P(I, D, I) + P(I, I, D) = (0.7)(0.3)(0.3) + (0.3)(0.7)(0.3) + (0.3)(0.3)(0.7) = 0.189.
P(x = 2) = P(D, D, I) + P(D, I, D) + P(I, D, D)
= (0.7)(0.7)(0.3) + (0.7)(0.3)(0.7) + (0.3)(0.7)(0.7) = 0.441.
P(x = 3) = P(D, D, D) = (0.7)(0.7)(0.7) = 0.343.
We display the probability distribution in the following table.
0 0.027
1 0.189
2 0.441
3 0.343
7.15 a. The table below lists all possible outcomes and the corresponding x values and probabilities.
Outcome x Probability
FFFF 0 (0.8)(0.8)(0.8)(0.8) = 0.4096
FFFS 1 (0.8)(0.8)(0.8)(0.2) = 0.1024
FFSF 1 (0.8)(0.8)(0.2)(0.8) = 0.1024
FSFF 1 (0.8)(0.2)(0.8)(0.8) = 0.1024
SFFF 1 (0.2)(0.8)(0.8)(0.8) = 0.1024
FFSS 2 (0.8)(0.8)(0.2)(0.2) = 0.0256
FSFS 2 (0.8)(0.2)(0.8)(0.2) = 0.0256
FSSF 2 (0.8)(0.2)(0.2)(0.8) = 0.0256
SFFS 2 (0.2)(0.8)(0.8)(0.2) = 0.0256
SFSF 2 (0.2)(0.8)(0.2)(0.8) = 0.0256
SSFF 2 (0.2)(0.2)(0.8)(0.8) = 0.0256
FSSS 3 (0.8)(0.2)(0.2)(0.2) = 0.0064
SFSS 3 (0.2)(0.8)(0.2)(0.2) = 0.0064
SSFS 3 (0.2)(0.2)(0.8)(0.2) = 0.0064
SSSF 3 (0.2)(0.2)(0.2)(0.8) = 0.0064
SSSS 4 (0.2)(0.2)(0.2)(0.2) = 0.0016
155
From this we deduce the following probability distribution of x.
0 0.4096
1 0.4096
2 0.1536
3 0.0256
4 0.0016
b. Both 0 and 1 are most likely values since each has probability 0.4096 of occurring.
c. P(at least 2 of the 4 have earthquake insurance) = P(x ≥ 2) = 0.1536 + 0.0256 + 0.0016 =
0.1808
7.16 Let the slips be numbered 1, 2, 3, 4, and 5, with corresponding dollar values of 1, 1, 1, 10, and 25.
Suppose the notation (1,4) means slips numbered 1 and 4 are selected. In this case the larger of the
two dollar values would be 10. The following table lists all possible outcomes, the corresponding
dollar values, and the values of w. Each outcome has a probability of occurrence equal to 1/10 = 0.1.
Outcome Dollar Values Value of w

(slips on the slips (in dollars)
selected)
(1,2) $1, $1 1
(1,3) $1, $1 1
(1,4) $1, $10 10
(1,5) $1, $25 25
(2,3) $1, $1 1
(2,4) $1, $10 10
(2,5) $1, $25 25
(3,4) $1, $10 10
(3,5) $1, $25 25
(4,5) $10, $25 25
Hence the probability distribution of w is as shown in the table below.
Value of w (in dollars) Probability

1 3/10 = 0.3
10 3/10 = 0.3
25 4/10 = 0.4
7.17 a. The smallest value of y is 1. The outcome corresponding to this is S.

The second smallest value of y is 2. The outcome that corresponds to this is FS.
b. The set of all possible y values is the set of all positive integers.
c. P(y = 1) = P(S) = 0.7

P(y = 2) = P(FS) = (0.3)(0.7) = 0.21
P(y = 3) = P(FFS) = (0.3)(0.3)(0.7) = (0.3)2 (0.7) = 0.063
P(y = 4) = P(FFFS) = (0.3)(0.3)(0.3)(0.7) = (0.3)3 (0.7) = 0.0189
P(y = 5) = P(FFFFS) = (0.3)(0.3)(0.3)(0.3)(0.7) = (0.3)4 (0.7) = 0.00567
We do see a pattern here. In fact,

p(y) = P(y-1 failures followed by a success) = (0.3)y-1 (0.7) for y = 1, 2, 3, ....
156
7.18 a. Because ∑ p ( y ) = 1 , we have k + 2k + 3k + 4k + 5k = 1 which implies that 15k =1. So k = 1/15.
1 2 3 6
b. P(at most 3 forms required ) = p(1) + p(2) + p(3) = + + = = 0.4.
15 15 15 15
c. P( between 2 and 4 forms (inclusive) will be required )
2 3 4 9
= p(2) + p(3) + p(4) = + + = = 0.6.
15 15 15 15
y2
d. With p( y ) =, the sum of the probabilities corresponding to y = 1, 2, 3, 4, and 5 is equal to
50
1 4 9 16 25 55
+ + + + = = 1.1 . Since the sum of the probabilities is not equal to 1,
50 50 50 50 50 50
y2
p( y ) = cannot be a probability distribution function for a random variable.
50
7.19 The following table lists all possible outcomes and the corresponding values of y.
Magazine Magazine Probability Value of y

1 2
arrives on arrives on
Wednesday Wednesda (0.4)(0.4) = 0.16 0
y
Wednesday Thursday (0.4)(0.3) = 0.12 1
Wednesday Friday (0.4)(0.2) = 0.08 2
Wednesday Saturday (0.4)(0.1) = 0.04 3
Thursday Wednesda (0.3)(0.4) = 0.12 1
y
Thursday Thursday (0.3)(0.3) = 0.09 1
Thursday Friday (0.3)(0.2) = 0.06 2
Thursday Saturday (0.3)(0.1) = 0.03 3
Friday Wednesda (0.2)(0.4) = 0.08 2
y
Friday Thursday (0.2)(0.3) = 0.06 2
Friday Friday (0.2)(0.2) = 0.04 2
Friday Saturday (0.2)(0.1) = 0.02 3
Saturday Wednesda (0.1)(0.4) = 0.04 3
y
Saturday Thursday (0.1)(0.3) = 0.03 3
Saturday Friday (0.1)(0.2) = 0.02 3
Saturday Saturday (0.1)(0.1) = 0.01 3
From the above table we deduce the following probability distribution for y.
Value of y Probability
0 0.16
1 0.33
2 0.32
3 0.19
157
7.20 a.
0 5 10 15 20 25 30 35 40 45 50
b.
0 5 10 15 20 25 30 35 40 45 50
c.
0 5 10 15 20 25 30 35 40 45 50
d.
0 5 10 15 20 25 30 35 40 45 50
158
e.
0 5 10 15 20 25 30 35 40 45 50
7.21 a. P(at most 5 minutes elapses before dismissal) = (1/10)(5 − 0) = 0.5
b. P(3 ≤ x ≤ 5) = (1/10)(5 − 3) = 0.2
7.22 i. P(2 < x < 3) = (0.1)(3 – 2) = 0.1

ii. P(2 ≤ x ≤ 3) = (0.1)(3 – 2) = 0.1
iii. P(x < 2) = (0.1)(2 – 0) = 0.2
iv. P(x > 7) = (0.1)(10 – 7) = 0.3
So, P(2 < x < 3) = P(2 ≤ x ≤ 3) < P(x < 2) < P(x > 7).
7.23 a. The density curve for x is shown below.
0.08
0 5 10 15 20 25
b. The area under the density curve between x = 7.5 and x = 20.0 must be 1. This implies that
the height of the density curve = 1/(20 – 7.5) = 1/12.5 = 0.08.
c. P(x is at most 12) = (0.08)(12 – 7.5) = (0.08)(4.5) = 0.36.
d. P(x is between 10 and 15) = (0.08)(15 – 10) = 0.4.

P(x is between 12 and 17) = (0.08)(17 – 12) = 0.4.
The two probabilities are equal because the distribution is uniform and therefore the
probabilities depend only on the length of the interval for which the probability is being
sought. This ensures they have the same area.
159
⎛ 2 + 1⎞
7.24 a. P(x < ½) = (0.5 – 0) ⎜ ⎟ =0.75.
⎝ 2 ⎠
⎛ 2 + 1⎞
b. P(x ≤ ½) = (0.5 – 0) ⎜ ⎟ =0.75.
⎝ 2 ⎠
⎛ 2 + 1.5 ⎞
c. P(x < ¼) = (0.25 – 0) ⎜ ⎟ = 7/16 = 0.4375.
⎝ 2 ⎠
d. P( ¼ < x < ½) = P(x < ½ ) – P(x < ¼ ) = 0.75 – 0.4375 = 0.3125.
e. P(sales exceed half a ton) = P(x > ½ ) = 1 – P(x ≤ ½ ) = 1 – 0.75 = 0.25.
f. P(sales is at least ¼ ton ) = P( x ≥ ¼ ) = 1 – P(x < ¼ ) = 1 – 0.4375 = 0.5625.
7.25 a. The height of the density curve must be 1/20 = 0.05. So

P(x is less than 10 minutes) = (0.05)(10 – 0) = 0.5.
P(x is more than 15 minutes) = (0.05)(20 – 15) = 0.25.
b. P(x is between 7 and 12 minutes) = (0.05)(12 – 7) = 0.25.
c. P(x < c) = (0.05)(c) = 0.9 when c = 0.9/0.05 = 18. So the value of c is 18.
7.26 a. area under the density curve = (0.5)(40)(0.05) = 1.
b. P(w < 20) = (0.5)(20)(0.05) = 0.5; P(w > 30) = (0.5)(10)(0.025) = 0.125.
c. P(w is between 10 and 30) = 1 - P(w is less than 10) - P(w is greater than 30)
= 1 – (0.5)(10 – 0)(0.25) – (0.5)(40 – 30)(0.25) = 1 – 0.125 – 0.125 = 0.75.
⎛ 0.5 + 1 ⎞
7.27 a. P(x ≤ 0.5) = 0.5 ⎜ ⎟ = 0.5(0.75) = 0.375
⎝ 2 ⎠
⎛ 0.75 + 1 ⎞
P(0.25 ≤ x ≤ 0.5) = 0.25 ⎜ ⎟ = 0.25(0.875) = 0.21875
⎝ 2 ⎠
⎛ 1.25 + 1.5 ⎞
P(x ≥ 0.75) = 0.25 ⎜ ⎟ = 0.25(1.375) = 0.34375
⎝ 2 ⎠
7 11
b. The mean of x is = 0.583333 and the standard deviation of x = = 0.276385. So,
12 144
P(x is more than one standard deviation from the mean value)
= P(x is less than (0.583333 – 0.276385) ) + P(x is greater than (0.583333 + 0.276385) )
= P(x is less than 0.306948) + P(x is greater than 0.859718)
⎛ 0.5 + 0.806948 ⎞ ⎛ 1.359718 + 1.5 ⎞
= (0.306948) ⎜ ⎟ + (1 – 0.859718) ⎜ ⎟ = 0.4012.
⎝ 2 ⎠ ⎝ 2 ⎠
160
7.28 a. μ x = (0)(0.54) + (1)(0.16) + (2)(0.06) + (3)(0.04) + (4)(0.20) = 1.2.
b. P(x > μ x ) = P(x > 1.2 ) = 0.06 + 0.04 + 0.20 = 0.30.
7.29 a. μ y = (0)(0.65) + (1)(0.20) + (2)(0.10) + (3)(0.04) + (4)(0.01) = 0.56. In the long run, the
average number of broken eggs per carton will equal 0.56.
b. P( number of broken eggs is less than μ y ) = P(y is less than 0.56) = P(y = 0) = 0.65. So, in
the long run, about 65% of the cartons will have fewer than μ y broken eggs. This is not
surprising because the distribution of y is skewed to the right and so the mean of y is greater
than the median of y. We would expect the required proportion to be more than 0.5 because
of the skewness.
c. The indicated calculation would be correct if the values of y were all equally likely, but this is
not the case. The values 0 and 1 occur more often than 2 or 3 or 4. Hence we need a
“weighted average” rather than a simple average.
7.30 z = number of unbroken eggs = 12 – number of broken eggs = 12 – y.

So μ z = 12 − μ y = 12 – 0.56 = 11.44.
7.31 σ x2 = (−1.2) 2 (0.54) + (−0.2) 2 (0.16) + (0.8) 2 (0.06) + (1.8) 2 (0.04) + (2.8) 2 (0.20) = 2.52. Hence,
σ x = 2.52 = 1.5875.
7.32 a. P(x is more than 1 standard deviation below its mean) = 0.02 + 0.03 + 0.09 = 0.14.
b. The values 1, 2 are the only values that are more than 2 standard deviations away from the
mean. So P(x is more than 2 standard deviations away from the mean) = 0.02 + 0.03 = 0.05.
7.33 a. μ x = (1)(0.05) + (2) (0.10) + (3)(0.12) + (4)(0.30) + (5)(0.30) + (6)(0.11) + (7)(0.01) + (8)(0.01)
= 4.12.
b. Using the definition of variance in the textbook, we get σ x2 = 1.94560 . So

σ x = 1.94560 = 1.3948 . The average squared distance of a value of x from its mean is
1.94560. The average distance of a value of x from its mean is approximately 1.3948.
c. P(x is within 1 standard deviation of the mean)

= P(x is between (4.12 – 1.3948) and (4.12 + 1.3948) )
= P( x is 3, 4 or 5) = 0.12 + 0.30 + 0.30 = 0.72.
d. P(x is more than 2 standard deviations from the mean) = P(x is 1, 7, or 8)

= 0.05 + 0.01 + 0.01 = 0.07.
7.34 a. μ x = (15)(0.1) + (30)(0.3) + (60)(0.6) = 46.5 (seconds)
b. Let y be the cost for a commercial. The possible values of y are 500, 800, 1000 (dollars). The
probability distribution of y is given by p(500) = 0.1, p(800) = 0.3, and p(1000) = 0.6. So the
average amount paid for commercials appearing on this station = μ y = (500)(0.1) +
(800)(0.3) + (1000)(0.6) = 890 (dollars).
161
7.35 μ x = 10,550. The author expects to get $ 10,550 under the royalty plan whereas the flat payment is
for $10,000. It would appear that the author should choose the royalty plan if he/she were quite
confident about his/her assessment of the probability distribution of x. On the other hand, P(x >
10,000) = 0.25 and P(x < 10000) = 0.35, so it is more likely that the royalty plan would yield an
amount less than 10,000 dollars than an amount greater than 10,000 dollars, so if the author isn’t
sure about his/her assessment of the probability distribution of x, then he/she might prefer the flat
payment plan.
7.36 Probability distribution 1:

x 1 2 3 4 5
p(x) 0.5 0 0 0 0.5
Probability distribution 2:
x 1 2 3 4 5
p(x) 0.2 0.2 0.2 0.2 0.2
The mean for distribution 1 is (1)(0.5) + (5)(0.5) = 3. The standard deviation is 3.9452.
The mean for distribution 2 is (1)(0.2) + (2)(0.2) + (3)(0.2) + (4)(0.2) + (5)(0.2) = 3. The standard
deviation is 1.41421.
Thus the two distributions have the same mean but quite different standard deviations.
7.37 a. y is a discrete random variable because there are only 6 possible values. Successive
possible values have gaps between them (this is always the case when the variable takes on
only a finite number of possible values).
b. P(paid more than $1.20 per gallon) = 0.10 + 0.16 + 0.08 + 0.06 = 0.40.
P(paid less than $1.40 per gallon) = 0.36 + 0.24 + 0.10 + 0.16 = 0.86.
c. The mean value of y is 126.808 (cents per gallon) and the standard deviation is 13.3162
(cents per gallon). In the long run, the average value of y will be 126.808 cents/gallon and the
deviation of y on any given day, from the mean value of y, will be about 13.3162 cents/gallon.
7.38 a. Mean value of x = 2.3.
b. Variance of x = 0.81 and standard deviation = 0.9.
7.39 The variable y is related to the variable x of Problem 7.38 by the relation y = 100 – 5x. Hence the
mean of y is 100 – 5(Mean of x) = 100 – (5)(2.3) = 88.5. The variance of y = (25)(variance of x) =
(25)(0.81) = 20.25.
7.40 a. Mean of x = 16.38 (in3) and variance = 3.9936 (in)2 .
b. Mean(Price) = 25 Mean(x) – 8.5 = (25)(16.38) – 8.5 = 401.00 (dollars).
c. Variance(Price) = (25)2 Variance(x) = (625)(3.9936) = 2496 (dollars)2 .

Standard deviation = 2496 = 49.96 .
162
7.41 a. Because if y > 0, ⇒ x2 > x1 ⇒ the diameter of the peg > the diameter of the hole and the peg
wouldn’t fit in the hole!
b. 0.253 – 0.25 = 0.003

c. 0.0022 + 0.0062 = 0.006
d. Yes, it is reasonable to assume they are independent, they are made by different tools and
are randomly selected.
e. With a standard deviation large than a mean, it seems fairly likely to obtain a negative value
of y so it would seem a relatively common occurrence to find a peg that was too big to fit in
the pre-drilled hole.
1 1 1 1
7.42 a. μ ( y ) = μ ( x1 −
x2 ) ⇒ μ ( y ) = μ ( x1 ) − μ ( x2 ) ⇒ μ ( y ) = μ ( x1 ) − μ ( x2 ) = 10 − (40) = 0
4 4 4 4
This would mean that the average score for those who know nothing and guess for every
question is zero – a just score!
b. It would be unreasonable because x1, the number if correct responses, and x2, the number of
incorrect responses are, of course not independent of each other.
7.43 a. Mean of x = 2.8; standard deviation of x = 1.289.
b. Mean of y = 0.7; standard deviation of y = 0.781.
c. Let w1 = total amount of money collected from cars. Then w1 = 3 x.

Mean of w1 = 3 (Mean of x) = (3)(2.8) = 8.4 dollars.
Variance of w1 = 9 (Variance of x) = (9)(1.66) = 14.94.
d. Let w2 = total amount of money collected from buses. Then w2 = 10 y.

Mean of w2 = 10 (Mean of y) = (10)(0.7) = 7 dollars.
Variance of w2 = 100 (Variance of y) = (100)(0.61) = 61.
e. Let z = total number of vehicles on the ferry. Then z = x + y.

Mean of z = Mean of x + Mean of y = 2.8 + 0.7 = 3.5.
Variance of z = Variance of x + Variance of y = 1.66 + 0.61 = 2.27.
f. w = w1 + w2 , so Mean of w = Mean of w1 + Mean of w2 = 8.4 + 7 = 15.4 dollars.

Variance of w = Variance of w1 + Variance of w2 = 14.94 + 61 = 75.94.
7.44 a. Mean of xR = 3.5; variance of xR = 2.9167; standard deviation of xR = 1.7078
b. Mean of xB = 3.5; variance of xB = 2.9167; standard deviation of xB = 1.7078
c. μ(w1) = μ(xR + xB – 7) = μ(xR) + μ(xB) – μ(7) = 3.5 + 3.5 – 7 = 0

σ2(w1) = σ2(xR + xB – 7) = σ2 (xR) + σ2 (xB) + σ2 (7) = 2.9167 + 2.9167 + 0 = 4.3833
standard deviation, σ(w1) = 2.094
d. μ(w2) = μ(3(xR-xB) = 3(μ(xR))) – 3(μ(xB))) = 3.5 – 3.5 = 0

σ2 (w2) = σ2 (3(xR-xB) = 9(σ2(xR) + σ(xB)) = 9(2.9167 + 2.9167) = 26.2503
standard deviation, σ(w2) = 5.124
e. It depends on how much risk I want to take. the variability in my winnings would be much
greater in game 2; I could win much more, but I could also lose much more!
163
7.45 a. There are exactly 6 such outcomes. They are SFFFFF, FSFFFF, FFSFFF, FFFSFF,
FFFFSF, FFFFFS.
b. In a binomial experiment consisting of 20 trials, the number of outcomes with exactly 10 S’s
⎛ 20 ⎞
is equal to ⎜ ⎟ = 184756. The number of outcomes with exactly 15 S’s is equal to
⎝ 10 ⎠
⎛ 20 ⎞
⎜ ⎟ = 15504. The number of outcomes with exactly 5 S’s is also equal to 15504 because
⎝ 15 ⎠
⎛ 20 ⎞ ⎛ 20 ⎞
⎜ ⎟ = ⎜ ⎟.
⎝ 15 ⎠ ⎝ 5 ⎠
⎛4⎞
7.46 a p(2) = ⎜ ⎟ π 2 (1 − π )4 −2 = (6)(0.9)2(0.1)2 = 0.0486.
⎝2⎠
b. p(4) = (0.9)4 = 0.6561.
c. p(x ≤ 3) = p(0) + p(1) + p(2) + p(3) = 0.0001 + 0.0036 + 0.0486 + 0.2916 = 0.3439.
⎛6⎞
7.47 a. p(4) = ⎜ ⎟ π 4 (1 − π )6 − 4 = (15)(0.8)4(0.2)2 = 0.24576. This means, in the long run, in samples of
⎝4⎠
6 passengers selected from passengers flying a long route, the proportion of the time exactly
4 out of the 6 will sleep or rest will be close to 0.24576.
b. p(6) = (0.8)6 = 0.262144.
c. p(x ≥ 4) = p(x = 4) + p(x = 5) + p(x = 6) = 0.245760 + 0.393216 + 0.262144 = 0.90112.
7.48 a. Using Appendix Table 9, we get p(8) = 0.302.
b. Using Appendix Table 9, P( x ≤ 7 ) = 1 – p(8) – p(9) – p(10) = 1 – 0.302 – 0.268 – 0.107 =

0.323.
c. P(more than half slept or rested) = p(6) + p(7) + p(8) + p(9) + p(10) = 0.088 + 0.201 + 0.302
+ 0.268 + 0.107 = 0.966.
⎛5⎞
7.49 a. p(2) = ⎜ ⎟ π 2 (1 − π )5 −2 = (10)(0.25)2(0.75)3 = 0.26367.
⎝ 2⎠
b. P(x ≤ 1) = p(0) + p(1) = 0.2373046875 + 0.3955078125 = 0.6328125.
c. P( 2 ≤ x) = 1 – P(x ≤ 1) = 1 – 0.6328125 = 0.3671875.
d. P( x ≠ 2 ) = 1 – P( x = 2 ) = 1 – p(2) = 1 – 0.26367 = 0.73633.
164
7.50
x = number of female p(x)
puppies in a litter of size 5
0 .03125
1 .15625
2 .31250
3 .31250
4 .15625
5 .03125
These probabilities can be obtained from Appendix Table 9 (the values are rounded to 3 decimal
places in that table). The probabilities may be calculated using the formula on page 337 of the
textbook.
⎛5⎞
For instance, p(2) = ⎜ ⎟ π 2 (1 − π )5 −2 = (10)(0.5)2(0.5)3 = 0.3125.
⎝ 2⎠
7.51 a. P(X = 10) = (0.85)10 = 0.1969
b. P(X ≤ 8) = 0.4557
c. p = 0.15, n = 500 mean = 75, st. dev. = 7.984
d. 25 is more than 3 standard deviations from the mean value of x, so yes, this is a surprising
result.
7.52 a. Let x = number of defective parts (successes). Then x has a binomial distribution with
n = 20 and π = 0.05. P(lot will be accepted) = p(0) + p(1) = 0.358 + 0.377 = 0.735.
b. Now n = 20 and π = 0.10. P(lot will be accepted) = p(0) + p(1) = 0.122 + 0.270 = 0.392.
c. Here n = 20 and π = 0.20. P(lot will be accepted) = p(0) + p(1) = 0.012 + 0.058 = 0.070.
7.53 Suppose the graphologist is just guessing, i.e., deciding which handwriting is which by simply a
coin toss. Then there is a 50% chance of guessing correctly in a single test. The probability of getting
6 or more correct in 10 trials, simply by guessing = p(6) + p(7) + p(8) + p(9) + p(10), where p(x) is the
probability that a binomial random variable with n = 10 and π = 0.5 will take the value x. Using
Appendix Table 9, we find this probability to be = 0.205 + 0.117 + 0.044 + 0.010 + 0.001 = 0.377.
Therefore, correctly guessing 6 or more out of 10 is not all that rare even without any special abilities.
So, the evidence given here is certainly not convincing enough to conclude that the graphologist has
any special abilities.
7.54 Assuming independence of the status of the different trees, we can use a binomial distribution with
n = 2000 and π = 0.1 as a reasonable model for x = number of damaged trees. The expected
number of trees showing damage = n π = (2000)(0.1) = 200. The standard deviation is nπ (1 − π ) =
2000(0.1)(1 − 0.1) = 13.4164.
7.55 a. P(at most 5 fail inspection) = p(0) + p(1) + p(2) + p(3) + p(4) + p(5), where p(x) is the
probability that a binomial random variable with n = 15 and π = 0.3 will take the value x.
Using Appendix Table 9, we get P(at most 5 fail inspection) = 0.005 + 0.030 + 0.092 + 0.170
+ 0.218 + 0.207 = 0.722.
b. P(between 5 and 10 (inclusive) fail inspection) = p(5) + p(6) + p(7) + p(8) + p(9) + p(10)
= 0.207+ 0.147 + 0.081 + 0.035 + 0.011 + 0.003 = 0.484.
165
c. Here, let x = number of cars that pass inspection. Then x is a binomial random variable
with n = 25 and π = 1 – 0.3 = 0.7. Hence the expected value of x is (25)(0.7) = 17.5 and the
standard deviation is 25(0.7)(1 − 0.7) = 2.2913.
7.56 a. x = number of correct answers. Here x is a binomial random variable with n = 100 and
π = 0.2 (1 in 5 chance of guessing correctly).
b. Expected score = (100)(0.2) = 20.
c. Variance = (100)(0.2)(0.8) = 16; Standard deviation = 4.
d. A score of 50 is 7.5 standard deviations away from the mean and the probability of this
1
occurring is at most equal to = 0.018 (use Chebyshev’s rule). Hence it is highly
(7.5)2
unlikely that a score of 50 can be obtained by purely guessing the answers for each question.
7.57 Here n/N = 1000/10000 = 0.1 which is greater than 0.05. So a binomial distribution is not a good
approximation for x = number of invalid signatures in a sample of size 1000 since the sampling is
done without replacement.
7.58 a. P(judge the coin to be biased | coin is fair ) = P( x ≤ 7 ) + P( x ≥ 18 ), where x is a

binomial random variable with n = 25 and π = 0.5. From Appendix Table 9, this probability is
calculated to be 0.044.
b. P(judge coin to be fair | π = 0.9) = P( 8 ≤ x ≤ 17 ) where x is a binomial random variable

with n = 25 and π = 0.9. From Appendix Table 9, we calculate this probability to be 0.002.
P(judge coin to be fair | π = 0.1) = P( 8 ≤ x ≤ 17 ) where x is a binomial random variable

with n = 25 and π = 0.1. From Appendix Table 9, we calculate this probability to be 0.002.
c. P(judge coin to be fair | π = 0.6) = P( 8 ≤ x ≤ 17 ) where x is a binomial random variable

with
n = 25 and π = 0.6. From Appendix Table 9, we calculate this probability to be 0.846.

with
The probabilities in this part are much larger than the probabilities in part b, because the true
value of π here is much closer to 0.5.
d. If the decision rule is changed as stated, then, for part a,

P(judge the coin to be biased |coin is fair ) = P( x ≤ 6 ) + P( x ≥ 19 ), where x is a binomial
random variable with n = 25 and π = 0.5. From Appendix Table 9, this probability is
calculated to be 0.010.

with
The probability of falsely “accusing” a fair coin to be biased is now lower. The price we pay is
that there is a greater probability of declaring a biased coin to be fair.
166
7.59 a. P(program is implemented | π = 0.8) = P(x ≤ 15) where x is a binomial random variable
with n = 25 and π = 0.8. Using Appendix Table 9, we calculate this probability to be 0.17.
b. P(program not implemented | π = 0.7) = P(x > 15) where x is a binomial random variable with
n = 25 and π = 0.7. Using Appendix Table 9, we calculate this probability to be 0.811.
P(program not implemented | π = 0.6 ) = P(x > 15) where x is a binomial random variable
c. Suppose the value 15 is changed to 14 in the decision criterion.
Then P(program is implemented | π = 0.8) = P(x ≤ 14) where x is a binomial random variable
with n = 25 and π = 0.8. Using Appendix Table 9, we calculate this probability to be 0.0.006.
P(program not implemented | π = 0.7) = P(x > 14) where x is a binomial random variable with
n = 25 and π = 0.7. Using Appendix Table 9, we calculate this probability to be 0.902.
P(program not implemented | π = 0.6 ) = P(x > 14) where x is a binomial random variable
The modified decision criterion leads to a lower probability of implementing the program
when it need not be implemented and a higher probability of not implementing the program
when it should be implemented.
7.60 a. Let x = number of registered voters in a random sample of size 25 who favor the ban.
Then x is a binomial random variable with n = 25 and π = 0.9. P( x > 20 ) = 0.902, using
Appendix Table 9.
b. P(at least 20 favor the ban) = P(exactly 20 favor the ban) + P( x > 20) = 0.065 + 0.902 (using
the result of part (a) = 0.967.
c. Mean value of x = (25)(0.9) = 22.5. Standard deviation = (25)(0.9)(1 − 0.9) = 1.5.
d. P(x < 20 | π = 0.9 ) = 1 – P(x ≥ 20) = 1 – 0.967 = 0.033. If indeed 90% or more of the
populace favors the ban, then there is only a 3.3% chance of finding less than 20 people
favoring the ban in a random sample of 25 people. This would certainly be considered to be
at odds with the assertion that at least 90% of the populace favors the ban.
7.61 a. Geometric distribution. We are not counting the number of successes in a fixed number
of trials; instead, we are counting the number of trials needed to achieve a single success.
b. P(exactly two tosses) = (0.9)(0.1) = 0.09.
c. P(more than three tosses will be required) = P(first three attempts are failures) = (0.9)3 =
0.729.
7.62 a. P(at most 2 boxes) = P(S) + P(FS) (here S means “Success” and F means “Failure”)
= 0.05 + (0.95)(0.05) = 0.0975.
b. P(exactly 4 boxes needed) = (0.95)3(0.05) = 0.04286875.
c. P(more than 4 boxes must be purchased) = P(first 4 boxes are failures) = (0.95)4
= 0.81450625.
7.63 a. Geometric distribution

b. P(X = 3) = 0.1084
c. P(X < 4) = P(X ≤ 3) = 0.3859
d. P(X > 3) = 1 – P(X ≤ 3) = 1 - 0.3859 = 0.6141
167
7.64 a. P(z < 1.75) = 0.9599
b. P(z < −0.68) = 0.2483
c. P(z > 1.20) = 1 − P(z ≤ 1.20) = 1 − 0.8849 = 0.1151
d. P(z > −2.82) = 1 − P(z ≤ −2.82) = 1 − 0.0024 = 0.9976
e. P(−2.22 < z < 0.53) = P(z < 0.53) − P(z < −2.22) = 0.7019 − 0.0132 = 0.6887
f. P(−1 < z < 1) = P(z < 1) − P(z < −1) = 0.8413 − 0.1587 = 0.6826
g. P(−4 < z < 4) = P(z < 4) − P(z < −4) ≈ 1 − 0 = 1
7.65 a. P(z < −1.28) = 0.1003
b. P(z > 1.28) = 1 − P(z ≤ 1.28) = 1 − 0.8997 = 0.1003
c. P(−1 < z < 2) = P(z < 2) − P(z < −1) = 0.9772 − 0.1587 = 0.8185
d. P(z > 0) = 1 − P(z ≤ 0) = 1 − 0.5 = 0.5
e. P(z > −5) = 1 − P(z ≤ −5) ≈ 1 − 0 = 1
f. P(−1.6 < z < 2.5) = P(z < 2.5) − P(z < −1.6) = 0.9938 − 0.0548 = 0.9390
g. P(z < 0.23) = 0.5910
7.66 a. 0.9909
b. 0.9909
c. 0.1093
d. P(1.14 < z < 3.35) = P(z < 3.35) − P(z ≤ 1.14) = 0.9996 − 0.8729 = 0.1267
e. P(−0.77 ≤ z ≤ −0.55) = P(z ≤ −0.55) − P(z < −0.77) = 0.2912 − 0.2206 = 0.0706
f. P(2 < z) = 1 − P(z ≤ 2) = 1 − 0.9772 = 0.0228
g. P(−3.38 ≤ z) = 1 − P(z < −3.38) = 1 − 0.0004 = 0.9996
h. P(z < 4.98) ≈ 1
7.67 a. P(z < 0.1) = 0.5398
b. P(z < −0.1) = 0.4602
c. P(0.40 < z < 0.85) = P(z < 0.85) − P(z < 0.4) = 0.8023 − 0.6554 = 0.1469
d. P(−0.85 < z < −0.40) = P(z < −0.4) − P(z < −0.85) = 0.3446 − 0.1977 = 0.1469
e. P(−0.40 < z < 0.85) = P(z < 0.85) − P(z < −0.4) = 0.8023 − 0.3446 = 0.4577
168
f. P(−1.25 < z) = 1 − P(z ≤ −1.25) = 1 − 0.1056 = 0.8944
g. P(z < −1.5 or z > 2.5) = P(z < −1.5) + 1 − P(z ≤ 2.5) = 0.0668 + 1 − 0.9938 = 0.0730
7.68 a. z* = −1.96
b. z* = −2.33
c. z* = −1.645 (or −1.64 or −1.65)
d. z* = 2.05
e. z* = 2.33
f. P(z > z* or z < −z*) = 0.20 ⇒ P(z > z*) = 0.10 ⇒ z* = 1.28
7.69 a. P(z > z*) = 0.03 ⇒ z* = 1.88
b. P(z > z*) = 0.01 ⇒ z* = 2.33
c. P(z < z*) = 0.04 ⇒ z* = −1.75
d. P(z < z*) = 0.10 ⇒ z* = −1.28
7.70 a. P(−z* < z < z*) = 0.95 ⇒ P(z > z*) = 0.025 ⇒ z* = 1.96
b. P(−z* < z < z*) = 0.90 ⇒ P(z > z*) = 0.05 ⇒ z* = 1.645
c. P(−z* < z < z*) = 0.98 ⇒ P(z > z*) = 0.01 ⇒ z* = 2.33
d. P(−z* < z < z*) = 0.92 ⇒ P(z > z*) = 0.04 ⇒ z* = 1.75
7.71 a. 91st percentile = 1.34

b. 77th percentile = 0.74
c. 50th percentile = 0
d. 9th percentile = −1.34
e. They are negatives of one another. The 100pth and 100(1−p)th percentiles will be negatives
of one another, because the z curve is symmetric about 0.
7.72 a. P(x < 5.0) = P(z < (5 − 5)/0.2) = P(z < 0 ) = 0.5
b. P(x < 5.4) = P(z < (5.4 − 5)/0.2) = P(z < 2) = 0.9772
c. P(x ≤ 5.4) = P(z ≤ (5.4 − 5)/0.2) = P(z ≤ 2) = 0.9772
d. P(4.6 < x < 5.2) = P((4.6 − 5)/0.2 < z < (5.2 − 5)/0.2) = P(−2 < z < 1)
= P(z < 1) − P(z < −2) = 0.8413 − 0.0228 = 0.8185
e. P(4.5 < x) = P((4.5 − 5)/0.2 < z) = P(−2.5 < z) = 1 − P(z ≤ −2.5) = 1 − 0.0062 = 0.9938
f. P(4.0 < x) = P((4 − 5)/0.2 < z) = P(−5 < z) = 1 − P(z ≤ −5) ≈ 1 − 0 = 1
169
4000 − 3432
7.73 a. P(x > 4000) = P(z > ) = P(z > 1.1784 ) = 0.1193
482
3000 − 3432 4000 − 3432
P(3000 ≤ x ≤ 4000) = P( ≤z≤ )
482 482
= P(-0.8963 ≤ z ≤ 1.1784) = P(z ≤ 1.1784) – P(z < -0.8963) = 0.8807 - 0.1851 = 0.6956
2000 − 3432 5000 − 3432

b. P(x < 2000 ) + P( x > 5000 ) = P(z < ) + P(z > )
482 482
= P(z < -2.97095) + P(z > 3.25311) = 0.0015 + 0.00057 = 0.00207
3175.2 − 3432
c. P(x > 7 lbs) = P(x > 7(453.6) grams) = P(x > 3175.2) = P(z > )
482
= P(z >-0.53278) = 0.70191
d. We find x1* and x2* such that P(x < x1* ) = 0.0005 and P(x > x2* ) = 0.0005. The most extreme
0.1% of all birthweights would then be characterized as weights less than x1* or weights
greater than x2*. P(x < x1* ) = P(z < z1* ) = 0.0005 implies that z1* = -3.2905. So x1* = μ +
z1*σ = 3432 + 482(-3.2905) = 1846 grams. P(x > x1* ) = P(z > z2* ) = 0.0005 implies that z2* =
3.2905. So x2* = μ + z2*σ = 3432 + 482(3.2905) = 5018 grams. Hence the most extreme
0.1% of all birthweights correspond to weights less than 1846 grams or weights greater than
5018 grams.
e. If x is a random variable with a normal distribution and a is a numerical constant (not equal
to 0) then y = ax also has a normal distribution. Furthermore, mean of y = a × (mean of x)
and standard deviation of y = a × (standard deviation of x). The birth weight distribution
would be normal with mean 7.5663 lbs and standard deviation 1.06263 lbs. The probability
that the birth weight is greater than 7 lbs. is still 0.7019.
Let y be the birthweights measured in pounds. Recalling that one pound = 453.6 grams, we
⎛ 1 ⎞ 1
have y = ⎜ ⎟ x, so a = . The distribution of y is normal with mean equal to
⎝ 453.6 ⎠ 453.6
3432/453.6 = 7.56614 pounds and standard deviation equal to 482/453.6 = 1.06261 pounds.
7 − 7.56614
So P(y > 7 lbs) = P(z > ) = P(z > –0.53278) = 0.70291. As expected, this is the
1.06261
same answer that we obtained in part c.
2.9 − 3 3.1 − 3
7.74 P(a cork meets specifications) = P(2.9 < x < 3.1) = P( <z< )
0.1 0.1
= P(−1 < z < 1) = 0.8413 − 0.1587 = 0.6826
Therefore, the proportion of corks produced by this machine that will be defective is
1 − 0.6826 = 0.3174.
7.75 For the second machine, P(that a cork doesn’t meet specifications) = 1 − P(2.9 ≤ x ≤ 3.1)
2.9 − 3.05 3.1 − 3.05
= 1 − P( ≤z≤ ) = 1 − P(−15 ≤ z ≤ 5 ) ≈ 1 − 1 = 0.
0.01 0.01
14.8 − 15
7.76 a. P(x ≤ 14.8) = P(z ≤ ) = P(z ≤ −2) = 0.0228
0.1
170
14.7 − 15 15.1 − 15
b. P(14.7 ≤ x ≤ 15.1) = P( ≤z≤ )
0.1 0.1
= P(−3 ≤ z ≤ 1) = 0.8413 − 0.0013 = 0.8400
c. P(x ≤ 15) = 0.5

P(both hold at most 15) = (0.5)(0.5) = 0.25
c − 120
7.77 = −1.28 ⇒ c = 120 − 1.28(20) = 94.4
120
Task times of 94.4 seconds or less qualify an individual for the training.
7.78 P(not acceptable) = 1 − P(0.496 ≤ x ≤ 0.504)

0.496 − 0.499 0.504 − 0.499
= 1 − P( ≤z≤ )
0.002 0.002
= 1 − P(−1.5 ≤ z ≤ 2.5) = 1 − (0.9938 − 0.0668) = 0.0730
60 − 60
7.79 a. P(x ≤ 60) = P(z ≤ ) = P(z ≤ 0) = 0.5
15
P(x < 60) = P(z < 0) = 0.5
45 − 60 90 − 60
b. P(45 < x < 90) = P( <z< )
15 15
= P(−1 < z < 2) = 0.9772 − 0.1587 = 0.8185
105 − 60
c. P(x ≥ 105) = P(z ≥ ) = P(z ≥ 3) = 1 − P(z < 3) = 1 − 0.9987 = 0.0013
15
The probability of a typist in this population having a net rate in excess of 105 is only 0.0013.
Hence it would be surprising if a randomly selected typist from this population had a net rate
in excess of 105.
75 − 60
d. P(x > 75) = P(z > ) = P(z > 1) = 1 − P(z ≤ 1) = 1 − 0.8413 = 0.1587
15
P(both exceed 75) = (0.1587)(0.1587) = 0.0252
e. P(z < z*) = 0.20 ⇒ z* = −0.84

x* = μ + z*σ ⇒ x* = 60 + (−0.84)(15) = 60 − 12.6 = 47.4
So typing speeds of 47.4 words or less per minute would qualify individuals for this training.
50 − 45
7.80 a. P(x >50) = P[z > ] = P(z > 1) = 1 - 0.8413 = 0.1587
5
The proportion of students at this university who would be unable to finish in the allotted time
is 0.1587.
b. P(x < x*) = 0.90, P(z < z*) = 0.90 ⇒ z* = 1.28

x* = 45 + 1.28(5) = 45 + 6.4 = 51.4
If it is desired that 90% of the students finish, then about 51.4 minutes should be
allowed.
c. P(z < z*) = 0.25 ⇒ z* = −0.67

x* = 45 + (−0.67)(5) = 45 − 3.35 = 41.65
About 25% of the students should finish in 41.65 minutes or less.
171
7.81 Since this plot appears to be very much like a straight line, it is reasonable to conclude that the
normal distribution provides an adequate description of the steam rate distribution.
7.82 Due to the very distinct curvature of this plot, one would conclude that the cadmium concentration
distribution is not normal. The plot suggests that the observations are coming from a skewed
distribution.
7.83
Since the graph exhibits a pattern substantially different from that of a straight line, one would
conclude that the distribution of the variable "component lifetime" cannot be adequately modeled by a
normal distribution. It is worthwhile noting that this “deviation from normality” could be due to the
single outlying value of 422.6.
7.84
Bearing Load Life (million revs)
500
400
300
200
100
0
-2 -1 0 1 2
Expected Normal Scores
Since the graph exhibits a pattern very similar to that of a straight line, one would conclude that the
distribution of the variable "load-life" could be adequately modeled by a normal distribution.
172
7.85
Although the graph follows a straight line pattern approximately, there is a distinct “kink” in the graph
at about the value 45 on the vertical axis. Points corresponding to DDT concentration less than 45
seem to follow one straight line pattern while those to the right of 45 seem to follow a different straight
line pattern. A normal distribution may not be an appropriate model for this population.
7.86
16.3
16.2
16.1
Diameter
16.0
15.9
15.8
15.7
15.6
-2 -1 0 1 2
Normal Score
Since the graph exhibits a pattern similar to that of a straight line, one would conclude that the
distribution could be adequately modeled by a normal distribution.
173
7.87 Histograms of the square root transformed data as well as the cube root transformed data are given
below. It appears that the histogram of the cube root transformed data is more symmetric than the
histogram of the square root transformed data. (However, keep in mind that the shapes of these
histograms are somewhat dependent on the choice of class intervals.)
4
Frequency
0.5 0.7 0.9 1.1 1.3 1.5 1.7 1.9 2.1 2.3
Square Root of y
4
Frequency
0.60 0.72 0.84 0.96 1.08 1.20 1.32 1.44 1.56 1.68
Cube Root of y
174
7.88 a.
class frequency relative frequency
0-<300 8 0.186047
300-<600 12 0.27907
600-<900 5 0.116279
900-<1200 5 0.116279
1200-<1500 4 0.093023
1500-<1800 2 0.046512
1800-<2100 4 0.093023
2100-<2400 1 0.023256
2400-<2700 2 0.046512
0.30
Relative Frequency
0.20
0.10
0 300 600 900 1200 1500 1800 2100 2400 2700

survival time (days)
b. The histogram above has a positive skew since it has a long upper tail.
c. Many of the commonly used statistical inference procedures are valid only when the sample
data are from a population distribution that is at least approximately normal. If such inference
methods are to be used to draw conclusions from the survival data of this problem then a
transformation should be considered so that, on the transformed scale, the data are
approximately normally distributed.
175
7.89 a. The required frequency distribution is given below.
class frequenc relative
y frequency
0-<100 22 0.22
100-<200 32 0.32
200-<300 26 0.26
300-<400 11 0.11
400-<500 4 0.04
500-<600 3 0.03
600-<700 1 0.01
700-<800 0 0
800-<900 1 0.01
b.
0.30
Relative Frequency
0.20
0.10
0 100 200 300 400 500 600 700 800 900

Number of Cycles to Breakage
c. Data were square root transformed and the corresponding relative frequency histogram is
shown below. Clearly the distribution of the transformed data is more symmetric than that of
the original data.
0.30
Relative Frequency
0.20
0.10
0 3 6 9 12 15 18 21 24 27 30
Square Root of Number of Cycles to Breakage
176
7.90 a.
0.45
0.40
Relative Frequency 0.35
0.30
0.25
0.20
0.15
0.10
0.05
0
0 50 100 150
Number of Purchases
b. Computations needed to construct a density histogram (we need a density histogram rather
than a relative frequency histogram because of unequal class widths) are given below.
Class Interval Frequency Rel.Frequenc Class Density

y Interval Rel. Frequency
Frequency Width =
= Class Width
2071
10 − < 20 904 0.436504 1.30986 0.333245
20 − < 30 500 0.241429 1.00509 0.240207
30 − < 40 258 0.124577 0.84733 0.147024
40 − < 50 167 0.080637 0.74651 0.108019
50 − < 60 94 0.045389 0.6749 0.067253
60 − < 70 56 0.02704 0.62063 0.043569
70 − < 80 26 0.012554 0.57767 0.021733
80 − < 90 20 0.009657 0.54256 0.017799
90 − < 100 13 0.006277 0.51317 0.012232
100 − < 110 9 0.004346 0.48809 0.008904
110 − < 120 7 0.00338 0.46636 0.007248
120 − < 130 6 0.002897 0.4473 0.006477
130 − < 140 6 0.002897 0.43041 0.006731
140 − < 150 3 0.001449 0.41529 0.003488
150 − < 160 0 0 0.40166 0
160 − < 170 2 0.000966 0.38929 0.002481
177
A histogram of these data using the square root scale is given in the figure below.
0.3
0.2
Density
0.1
0.0
2 4 6 8 10 12 14
Square Root of Number of Purchases
This histogram is less skewed than the one in part a, but it is still far from being symmetric.
7.91 a.
0 56667889999
1 000011122456679999
2 0011111368
3 11246
4 18
5
6 8
7
8
9 39 HI: 448
b. A density histogram of the body mass data is shown below. It is positively skewed. The scale
on the horizontal axis extends upto body mass = 500 in order to accommodate the HI value
of 448.
0.045
0.040
0.035
0.030
Density
0.025
0.020
0.015
0.010
0.005
0.000
0 50 100 150 200 250 300 350 400 450 500

Body Mass
178
If statistical inference procedures based on normality assumptions are to be used to draw
conclusions from the body mass data of this problem, then a transformation should be
considered so that, on the transformed scale, the data are approximately normally
distributed.
c.
The normal curve fits the histogram of the log transformed data better than the histogram of
the original data.
d.
The histogram in this transformed scale does appear to have an approximate bell-shape, i.e.,
in the transformed scale, the data do appear to follow a normal distribution model.
179
7.92 The transformation appears to have been more successful for Mn than for Cu and Zn. The histogram
of log(Zn) shows a hint of bimodality. In all three cases the histograms appear to be more symmetric
in the transformed scale than in the original scale.
⎛ 100 − 0.5 − 100 100 + 0.5 − 100 ⎞

7.93 a. P( x = 100 ) = P ⎜ ≤z≤ ⎟ = P( – 0.03333 ≤ z ≤ 0.03333 )
⎝ 15 15 ⎠
= 0.02659.
⎛ 110 + 0.5 − 100 ⎞

b. P(x ≤ 110 ) = P ⎜ z ≤ ⎟ = P( z ≤ 0.7 ) = .75804.
⎝ 15 ⎠
⎛ 109 + 0.5 − 100 ⎞

c. P(x < 110 ) = P( x ≤ 109 ) = P ⎜ z ≤ ⎟ = P(z ≤ 0.6333) = 0.73674.
⎝ 15 ⎠
⎛ 75 − 0.5 − 100 125 + 0.5 − 100 ⎞

d. P( 75 ≤ x ≤ 125) = P ⎜ ≤z≤ ⎟ = P(– 1.7 ≤ z ≤ 1.7 ) = 0.9109.
⎝ 15 15 ⎠
⎛ 120 + 0.5 − 150 ⎞

7.94 a. P(x ≤ 120 ) = P ⎜ z ≤ ⎟ = P( z ≤ –2.95 ) = 0.00159.
⎝ 10 ⎠
⎛ 125 − 0.5 − 150 ⎞

b. P(125 ≤ x ) = P ⎜ ≤ z ⎟ = P( –2.55 ≤ z ) = 0.9946.
⎝ 10 ⎠
⎛ 135 − 0.5 − 150 160 + 0.5 − 150 ⎞

c. P( 135 ≤ x ≤ 160 ) = P ⎜ ≤z≤ ⎟ = P(–1.55 ≤ z ≤ 1.05 ) =
⎝ 10 10 ⎠
0.7926.
⎛ 650 − 0.5 − 500 ⎞

7.95 a. P( 650 ≤ x ) = P ⎜ ≤ z ⎟ = P( 1.9933 ≤ z ) = 0.02311.
⎝ 75 ⎠
⎛ 401 − 0.5 − 500 549 + 0.5 − 500 ⎞

b. P(400 < x < 550 ) = P( 401 ≤ x ≤ 549 ) = P ⎜ ≤z≤ ⎟
⎝ 75 75 ⎠
= P(-1.3267 ≤ z ≤ 0.66 ) = 0.6531.
⎛ 400 − 0.5 − 500 550 + 0.5 − 500 ⎞

c. P( 400 ≤ x ≤ 550 ) = P ⎜ ≤z≤ ⎟
⎝ 75 75 ⎠
= P(-1.34 ≤ z ≤ 0.6733 ) = 0.6595.
⎛ 30 − 0.5 − 30 30 + 0.5 − 30 ⎞
7.96 a. P( x = 30 ) = P ⎜ ≤z≤ = P( –0.1443 ≤ z ≤ 0.1443 ) = 0.1147.
⎝ 3.4641 3.4641 ⎟⎠
⎛ 25 − 0.5 − 30 25 + 0.5 − 30 ⎞
b. P( x = 25 ) = P ⎜ ≤z≤ = P(–1.5877 ≤ z ≤ –1.2990 ) = 0.04079.
⎝ 3.4641 3.4641 ⎟⎠
⎛ 25 + 0.5 − 30 ⎞
c. P( x ≤ 25 ) = P ⎜ z ≤ = P( z ≤ –1.2990 ) = 0.09697.
⎝ 3.4641 ⎟⎠
⎛ 25 − 0.5 − 30 40 + 0.5 − 30 ⎞
d. P( 25 ≤ x ≤ 40 ) = P ⎜ ≤z≤ = P(–1.5877 ≤ z ≤ 3.0311 ) =
⎝ 3.4641 3.4641 ⎟⎠
0.9426.
180
⎛ 26 − 0.5 − 30 39 + 0.5 − 30 ⎞
e. P( 25 < x < 40 ) = P(26 ≤ x ≤ 39 ) = P ⎜ ≤z≤
⎝ 3.4641 3.4641 ⎟⎠
= P(–1.2990 ≤ z ≤ 2.7424 ) = 0.89998.
7.97 Let x = number of mountain bikes in a sample of 100. Then x has a binomial distribution with n = 100
and π = 0.7. Its mean is equal to 70 and standard deviation is equal to 100(0.7)(0.3) = 4.58258.
⎛ 75 + 0.5 − 70 ⎞
a. P(at most 75 mountain bikes) = P(x ≤ 75 ) = P ⎜ z ≤ = P( z ≤ 1.2002 ) =
⎝ 4.58258 ⎟⎠
0.88497.
⎛ 60 − 0.5 − 70 75 + 0.5 − 70 ⎞
b. P( 60 ≤ x ≤ 75 ) = P ⎜ ≤z≤ = P(–2.29129 ≤ z ≤ 1.2002 ) =
⎝ 4.58258 4.58258 ⎟⎠
0.8740.
⎛ 81 − 0.5 − 70 ⎞
c. P( 80 < x ) = P( 81 ≤ x ) = P ⎜ ≤ z ⎟ = P(2.29129 ≤ z ) = 0.01097.
⎝ 4.58258 ⎠
d. P(at most 30 are not mountain bikes) = P(at least 70 are mountain bikes) = P( 70 ≤ x ) =
⎛ 70 − 0.5 − 70 ⎞
P⎜ ≤ z ⎟ = P( – 0.1091 ≤ z ) = 0.54344.
⎝ 4.58258 ⎠
7.98 Let x = number of false alarms in a sample of size 100. Then x is a binomial random variable with n =
100 and π = 0.25. Its mean is equal to (100)(0.25) = 25 and standard deviation is equal to
100(0.25)(0.75) = 4.33013.
⎛ 20 − 0.5 − 25 30 + 0.5 − 25 ⎞
a. P( 20 ≤ x ≤ 30)= P ⎜ ≤z≤ =P(–1.27017 ≤ z ≤ 1.27017 ) =
⎝ 4.33013 4.33013 ⎟⎠
0.79598.
⎛ 21 − 0.5 − 25 29 + 0.5 − 25 ⎞
b. P( 20 < x < 30 ) = P( 21 ≤ x ≤ 29 )= P ⎜ ≤z≤
⎝ 4.33013 4.33013 ⎟⎠
= P(–1.03923 ≤ z ≤ 1.03923 ) = 0.7013.
⎛ 35 − 0.5 − 25 ⎞
c. P( 35 ≤ x ) = P ⎜ ≤ z ⎟ = P( 2.19393 ≤ z ) = 0.01412.
⎝ 4.33013 ⎠
d. P(x is farther than 2 standard deviations from its mean ) = P( x ≤ 16 OR 34 ≤ x )
⎛ 16 + 0.5 − 25 ⎞ ⎛ 34 − 0.5 − 25 ⎞
= P(x ≤ 16 ) + P( 34 ≤ x ) = P ⎜ z ≤ ⎟ + P⎜ ≤ z⎟
⎝ 4.33013 ⎠ ⎝ 4.33013 ⎠
= P( – 1.96299 ≤ z ) + P(1.96299 ≤ z ) = 0.02482 + 0.02482 = 0.04964.
7.99 Let x = number of voters in a random sample of size 225 that favors a 7-day waiting period. Then x is
a binomial random variable with n = 225 and π = 0.65 . Its mean is 146.25 and standard deviation is
7.15454.
⎛ 150 − 0.5 − 146.25 ⎞

a. P( 150 ≤ x ) = P ⎜ ≤ z ⎟ = P( 0.45426 ≤ z ) = 0.32482.
⎝ 7.15454 ⎠
⎛ 151 − 0.5 − 146.25 ⎞
b. P( 150 < x ) = P( 151 ≤ x ) = P ⎜ ≤ z ⎟ = P(.59403 ≤ z ) = 0.27625.
⎝ 7.15454 ⎠
⎛ 124 + 0.5 − 146.25 ⎞
c. P( x < 125 ) = P( x ≤ 124 ) = ⎜ z ≤ ⎟ = P( z ≤ – 3.04003 ) = 0.001183.
⎝ 7.15454 ⎠
181
7.100 a. Here n = 50 and π = 0.05 . Since n π = 2.5 is less than 10, a normal approximation would not
be satisfactory.
b. Here n = 500 and π = 0.05 . So n π = 25 ≥ 10. A binomial approximation may be used.

If x = number of defective light bulbs in a random sample of 500, then x is binomial with n =
500 and π = 0.05 . Its mean is 25 and standard deviation is 4.8734.
⎛ 20 − 0.5 − 25 ⎞
So P( 20 ≤ x ) = P ⎜ ≤ z ⎟ = P( – 1.12858 ≤ z ) = 0.87046.
⎝ 4.8734 ⎠
7.101 a. Let x = number of mufflers replaced under the warranty in a random sample of 400
purchases. Then x is binomial with n = 400 and π = 0.2 Its mean is 80 and standard
deviation is 8.
⎛ 75 − 0.5 − 80 100 + 0.5 − 80 ⎞
P( 75 ≤ x ≤ 100 )= P ⎜ ≤z≤ ⎟ =P(–0.6875 ≤ z ≤ 2.5625 ) =
⎝ 8 8 ⎠
0.74892.
⎛ 70 + 0.5 − 80 ⎞
b. P(x ≤ 70 ) = P ⎜ z ≤ ⎟ = P( z ≤ – 1.1875 ) = 0.11752.
⎝ 8 ⎠
⎛ 49 + 0.5 − 80 ⎞
c. P( x < 50 ) = P(x ≤ 49 ) = P ⎜ z ≤ ⎟ = P( z ≤ – 3.8125 ) = 0.00006878. This
⎝ 8 ⎠
probability is so close to zero that, if indeed the 20% figure were true, it is highly unlikely that
fewer than 50 mufflers among the 400 randomly selected purchases were ever replaced.
Therefore the 20% figure is highly suspect.
Exercises 7.102 – 7.124
7.102 a. The proportion of the population that is sampled is 5000/40000 = 1/8 = 0.125 = 12.5%. Since
the proportion sampled is greater than 5% and sampling is done without replacement, a
binomial distribution would not provide a good approximation for calculating probabilities. So
we cannot use the methods of this chapter to approximate the required probability.
b. Here a binomial distribution may be used. Let x = number of ASTA members in a random
sample of 100 agencies. Then x is binomial with n = 100 and π = 11000/40000 = 0.275. Its
mean is 27.5 and standard deviation is 4.46514.
c. No. The variance doubles, so the standard deviation increases by a factor of 2.
7.103 Let x = number of customers out of the 15 who want the diet Coke. Then x is binomial with n = 15 and
π = 0.6. P(each of the 15 is able to purchase the drink desired) = P( 5 ≤ x ≤ 10 ) = 0.025 + 0.061 +
0.118 + 0.177 + 0.207 + 0.196 = 0.784 (using Appendix Table 9).
7.104 a. P( x ≤ 3 ) = 0.70.
b. P( x < 3 ) = 0.45.
c. P( 3 ≤ x ) = 0.55.
d. P( 2 ≤ x ≤ 5 ) = 0.71.
e. P( x = 2 or 3 or 4 ) = 0.65. (Note that, if 2 lines are not in use, then 4 are in use; if 3 are not
in use, then 3 are in use; if 4 are not in use, then 2 are in use.)
f. P( x ≤ 2 ) = 0.45.
182
7.105 a. Mean of x = 2.64000; standard deviation = 1.53961.
b. P( x is farther than 3 standard deviations from its mean )

= P( x is less than -1.97883 OR x is greater than 7.25883 ) = 0.
7.106 a. P( y = 2 ) = P( first two batteries tested are satisfactory ) = (0.8)(0.8) = 0.64.
b. P( y = 3 ) = P( USS ) + P(SUS) = (0.2)(0.8)(0.8) + (0.8)(0.2)(0.8) = 0.256.
c. In order for y to take the value 5, there must be only 1 satisfactory battery among the first four
batteries selected and the fifth battery selected must be a satisfactory one. The four
outcomes that satisfy this requirement are SUUUS, USUUS, UUSUS, UUUSS. Hence p(5) =
(0.8)(0.2)(0.2)(0.2)(0.8) + (0.2)(0.8)(0.2)(0.2)(0.8) + (0.2)(0.2)(0.8)(0.2)(0.8) +
(0.2)(0.2)(0.2)(0.8)(0.8) = 0.02048.
d. In order that exactly y batteries are tested to find two satisfactory batteries, there must be
exactly one satisfactory battery among the first y-1 batteries tested and the yth battery tested
must be a satisfactory one. The probability of having exactly one satisfactory battery among
⎛ y − 1⎞ 1 y −1
the first y-1 is given by the binomial formula. It is equal to ⎜ ⎟ π (1 − π ) . The probability
⎝ 1 ⎠
for the last battery to be satisfactory is π . By independence of the outcomes of individual
⎛ y − 1⎞ 1 y −1
batteries, the required probability is the product of ⎜ ⎟ π (1 − π ) and π , i.e.,
⎝ 1 ⎠
⎛ y − 1⎞ 2 y −1
p( y ) = ⎜ ⎟ π (1 − π ) = ( y − 1)π 2 (1 − π )y −1 where π = 0.8 and y = 2,3,4,.......
⎝ 1 ⎠
⎛ 0.525 − 0.5 0.575 − 0.5 ⎞
7.107 a. P(0.525 ≤ y ≤ 0.550 ) = P ⎜ ≤z≤ = P( 1 ≤ z ≤ 2 ) = 0.1359.
⎝ 0.025 0.025 ⎟⎠
b. P( y exceeds its mean value by more than 2 standard deviations) = P( 2 < z ) = 0.02275.
c. Let π = P(a randomly selected pizza has at least 0.475 lb of cheese) = P( 0.475 ≤ y )
P( – 1 ≤ z ) = 0.84134. Let x = number of pizzas in a random sample of size 3 that have at
least 0.475 lb of cheese. Then x is a binomial random variable with n = 3 and π = 0.84134.
Therefore the probability that all 3 chosen pizzas have at least 0.475 lb of cheese = P( x = 3 )
= (0.84134)3 = 0.59555.
⎛ 29 − 30 31 − 30 ⎞
7.108 a. P( 29 ≤ y ≤ 31 ) = P ⎜ ≤z≤ = P( – 0.83333 ≤ z ≤ 0.83333 ) = 0.59534.
⎝ 1.2 1.2 ⎟⎠
⎛ 25 − 30 ⎞
b. P(y < 25 ) = P ⎜ z < = P( z < – 4.1667 ) = 0.00001545. This probability is so close to
⎝ 1.2 ⎟⎠
zero that one wouldn’t at all expect a car selected from this model type to have a fuel
efficiency rating of less than 25 mpg. It would definitely be a surprise.
c. Probability of a randomly selected car having an efficiency exceeding 32 mpg = P( y > 32 ) =

⎛ 32 − 30 ⎞
P⎜ < z ⎟ = P( 1.6667 < z ) = 0.04779. Hence, the probability that all three cars in a
⎝ 1.2 ⎠
random sample of size 3 will have fuel efficiencies exceeding 32 mpg = (0.04779)3 =
0.0001091.
⎛ c − 30.0 ⎞ c − 30.0
d. We want c such that P ⎜ ⎟ > 0.95. This occur when > −1.645 , or c > 28.026
⎝ 1.2 ⎠ 1.2
183
7.109 a. P( y > 45 ) = P( z > –1.5 ) = 0.9332.
b. Let K denote the amount exceeded by only 10% of all clients at a first meeting.
⎛ K − 60 ⎞ K − 60
Then 0.10 = P( y > K )=P ⎜ z > ⎟ implies that = 1.2816. So K = 72.816 minutes.
⎝ 10 ⎠ 10
⎛ y ⎞ ⎛ Mean of y ⎞
c. Let R = revenue. Then R = 10 + 50 ⎜ ⎟ . So mean value of R = 10 + 50 ⎜ ⎟=
⎝ 60 ⎠ ⎝ 60 ⎠
⎛ 60 ⎞
10 + 50 ⎜ ⎟ = 60 dollars.
⎝ 60 ⎠
7.110 a. P(player functions for at least 4 hours) = P(both batteries function for at least 4 hours). The
⎛ 4−6⎞
probability that a randomly chosen battery will function for at least 4 hours = P ⎜ z >
⎝ 0.8 ⎟⎠
= P( z > –2.5 ) = 0.99379. Hence, P(both batteries function for at least 4 hours) = (0.99379)2
= 0.98762.
b. The probability that a randomly chosen battery will function for more than 7 hours =
⎛ 7−6⎞
P⎜z > = P( z > 1.25 ) = 0.10565. So, P(both batteries function for more than 7 hours)
⎝ 0.8 ⎟⎠
= (0.10565)2 = 0.0112. Finally, P(player works for at most seven hours) = 1 – P(player works
for more than 7 hours) = 1 – P(both batteries work for more than 7 hours) = 1 – 0.0112 =
0.9888.
c. P( player will work for more than z* hours ) = P(each battery works for more than z* hours ) =
2
⎛ z* − 6 ⎞
P(a randomly chosen battery works for more than z* hours )2 = P ⎜ z > ⎟ = 0.05 implies
⎜ 0.8 ⎟⎠
⎝
⎛ z* − 6 ⎞ z* − 6
that P ⎜ z > ⎟= 0.05 = 0.22361. So, = 0.76 (using Appendix Table 2 ).
⎜ 0.8 ⎟⎠ 0.8
⎝
So z* = 6.608.
4.9 − 5 5.2 − 5
7.111 P(x < 4.9) = P(z < ) = P(z < −2) = 0.0228. P(5.2 < x) = P( < z) = P(4 < z) ≈ 0
0.05 0.05
3−4 5−4
7.112 a. P(3 < x < 5) = P( <z< ) = P(−1 < z < 1)
1 1
= P(z < 1) − P(z < −1) = 0.8413 − 0.1587 = 0.6826
4−4
b. P(x < 4) = P(z < ) = P(z < 0) = 0.5
1
7−4
c. P(7 < x) = P( < z) = P(3 < z) = 1 − P(z ≤ 3) = 1 − 0.9987 = 0.0013
1
d. The largest 10% of moisture loss values are those which exceed the 90th percentile.
The 90th percentile is 4 + 1.28(1) = 5.28.
e. P(x < 3) + P(x > 5) = 1 – P(3 < x < 5) = 1 - .6826 = .3174.
184
67 − 66
7.113 a. P(x < 5'7") = P(x < 67") = P(z < ) = P(z < 0.5) = 0.6915
2
No, the claim that 94% of all women are shorter than 5'7" is not correct. Only about 69% of
all women are shorter than 5'7".
b. About 69% of adult women would be excluded from employment due to the height
requirement.
7.114 a. Assume π = 0.5. The outcomes of the binomial experiment under consideration, the
corresponding values of y, and the probabilities of the outcomes are listed in the table below.
Outcome y p Outcome y p Outcome y p Outcome y p

SSSS 4 1/16 FSSS 3 1/16 FSSF 2 1/16 FSFF 1 1/16
SSSF 3 1/16 SSFF 2 1/16 FSFS 1 1/16 FFSF 1 1/16
SSFS 2 1/16 SFSF 1 1/16 FFSS 2 1/16 FFFS 1 1/16
SFSS 2 1/16 SFFS 1 1/16 SFFF 1 1/16 FFFF 0 1/16
The probability distribution of y is given below. Its mean is 1.6875.
y p(y)
0 1/16
1 7/16
2 5/16
3 2/16
4 1/16
b. Here π = 0.6. The outcomes of the binomial experiment under consideration, the
corresponding values of y, and the probabilities of the outcomes are listed in the table below.
Outcome p Outcome y p Outcome y p Outcome y p

SSSS 4 0.1296 FSSS 3 0.0864 FSSF 2 0.0576 FSFF 1 0.0384
SSSF 3 0.0864 SSFF 2 0.0576 FSFS 1 0.0576 FFSF 1 0.0384
SSFS 2 0.0864 SFSF 1 0.0576 FFSS 2 0.0576 FFFS 1 0.0384
SFSS 2 0.0864 SFFS 1 0.0576 SFFF 1 0.0384 FFFF 0 0.0256
y p(y)
0 0.0256
1 0.3264
2 0.3456
3 0.1728
4 0.1296
c. Assume π = 0.5. The outcomes of the binomial experiment under consideration, the
corresponding values of z, and the probabilities of the outcomes are listed in the table below.
Outcome z p Outcome z p Outcome z p Outcome z p

SSSS 4 1/16 FSSS 3 1/16 FSSF 2 1/16 FSFF 2 1/16
SSSF 3 1/16 SSFF 2 1/16 FSFS 1 1/16 FFSF 2 1/16
SSFS 2 1/16 SFSF 1 1/16 FFSS 2 1/16 FFFS 3 1/16
SFSS 2 1/16 SFFS 2 1/16 SFFF 3 1/16 FFFF 4 1/16
185
z p(z)
1 2/16
2 8/16
3 4/16
4 2/16
7.115 a. The 36 possible outcomes and the corresponding values of w are listed in the following table.
The notation (i, j) means that Allison arrived at time i P.M and Teri arrived at time j P.M.
Outcom w Outcom w Outcom w Outcom w

e e e e
(1,1) 0 (2,4) 2 (4,1) 3 (5,4) 1
(1,2) 1 (2,5) 3 (4,2) 2 (5,5) 0
(1,3) 2 (2,6) 4 (4,3) 1 (5,6) 1
(1,4) 3 (3,1) 2 (4,4) 0 (6,1) 5
(1,5) 4 (3,2) 1 (4,5) 1 (6,2) 4
(1,6) 5 (3,3) 0 (4,6) 2 (6,3) 3
(2,1) 1 (3,4) 1 (5,1) 4 (6,4) 2
(2,2) 0 (3,5) 2 (5,2) 3 (6,5) 1
(2,3) 1 (3,6) 3 (5,3) 2 (6,6) 0
From the above table we deduce the following probability distribution of w.
w p(w)
0 6/36
1 10/36
2 8/36
3 6/36
4 4/36
5 2/36
b. The expected value of w is equal to 70/36 = 1.9444 hours.
7.116 The order in which the four people are tested and the corresponding values of x are given in the table
below. Every outcome listed in the table has probability equal to 1/24.
Outcom x Outcom x Outcom x Outcom x

e e e e
a,b,c,d 1 b,a,c,d 1 c,a,b,d 2 d,a,b,c 2
a,b,d,c 1 b,a,d,c 1 c,a,d,b 2 d,a,c,b 2
a,c,b,d 1 b,c,a,d 1 c,b,a,d 2 d,b,a,c 2
a,c,d,b 1 b,c,d,a 1 c,b,d,a 2 d,b,c,a 2
a,d,b,c 1 b,d,a,c 1 c,d,a,b 3 d,c,a,b 3
a,d,c,b 1 b,d,c,a 1 c,d,b,a 3 d,c,b,a 3
From the above table we deduce the following probability distribution of x.
x p(x)
1 12/24
2 8/24
3 4/24
186
7.117 a. We use L to denote that Lygia won a game and B to denote Bob won a game. By a sequence
such as BLLB we mean that Bob won the first and the fourth games and Lygia won the
second and the third games. Then x = 4 occurs if either BBBB or LLLL occurs. So p(4) =
(0.6)4 + (0.4)4 = 0.1552.
b. The outcomes that lead to x = 5 are, LLLBL, LLBLL, LBLLL, BLLLL, BBBLB, BBLBB, BLBBB,
LBBBB. There for p(5) = 4 (0.6)4(0.4) + 4 (0.6)(0.4)4 = 0.2688.
c. The maximum value of x is 7 and the minimum value of x is 4. The following table gives the
distribution of x.
x p(x)
4 (0.6) + (0.4)4 = 0.1552
4
5 ⎛ 4⎞ ⎛ 4⎞
⎜ ⎟ (0.6) (0.4) + ⎜ ⎟ (0.6)(0.4) = 0.26880
4 4
⎝3⎠ ⎝3⎠
6 ⎛5⎞ ⎛5⎞
⎜ ⎟ (0.6) (0.4) + ⎜ ⎟ (0.6) (0.4) = 0.29952
4 2 2 4
⎝3⎠ ⎝3⎠
7 ⎛6⎞ ⎛6⎞
⎜ ⎟ (0.6) (0.4) + ⎜ ⎟ (0.6) (0.4) = 0.27648
4 3 3 4
⎝ ⎠
3 ⎝ ⎠
3
d. The expected value of x is equal to (4)(0.1552)+(5)(0.2688) + (6)(0.29952) + (7)(0.27648) =

5.69728.
7.118 The probability distribution of y is given in the following table. Note that if the play ends after x games,
then the number of games won by the loser is x – 4. Thus P(y=0) = P(x=4), P(y=1)=P(x=5), etc.
y p(y)
0 0.1552
1 0.26880
2 0.29952
3 0.27648
7.119 x = number among three randomly selected customers who buy brand W. In the following, we use
the sequence such as DWP to mean that the first customer buys brand D, the second buys brand W,
and the third buys brand P, etc.
P(x = 0) = P(each of the three customers failed to buy brand W) = (1– 0.4)3 = 0.216.
P( x = 1 ) = P(exactly one of 3 customers bought brand W) = 3(0.4)(0.6)2 = 0.432.
P( x = 2 ) = P(exactly two of 3 customers bought brand W) = 3(0.4)2(0.6) = 0.288.
P( x = 3 ) = P(all three bought brand W) = (0.4)3 = 0.064.
89 − 78
7. 120 P(89 ≤ x) = P( ≤ z) = P(1.57 ≤ z) = 1 − P(z < 1.57) = 1 − 0.9418 = 0.0582.
7
Since you were in the upper 5.82%, you did get an A.
5.9 − 6 6.15 − 6
7.121 a. P(5.9 < x < 6.15) = P( <z< ) = P(−1 < z < 1.5)
0.1 0.1
= 0.9332 − 0.1587 = 0.7745
6.1 − 6
b. P(6.1 < x) = P( < z) = P(1 < z) = 1 − P(z ≤ 1) = 1 − 0.8413 = 0.1587
0.1
187
5.95 − 6
c. P(x < 5.95) = P(z < ) = P(z < −0.5) = 0.3085
0.1
d. The largest 5% of the pH values are those pH values which exceed the 95th percentile.
The 95th percentile is 6 + 1.645(0.1) = 6.1645.
7.122 20th percentile = 700 + (−0.84)(50) = 658

They should replace the bulbs after about 658 hours.
7.123 a. Let x= number among the 200 who are uninsured. Then x is binomial with n= 200 and
π = 0.16. Its mean is (200)(0.16) = 32 and standard deviation is (200)(0.16)(1 − 0.16) =
5.18459.
⎛ 25 − 0.5 − 32 40 + 0.5 − 32 ⎞
b. P( 25 ≤ x ≤ 40 ) = P ⎜ ≤z≤ = P( –1.44659 ≤ z ≤ 1.63947 ) =
⎝ 5.18459 5.18459 ⎟⎠
0.87544.
⎛ 51 − 0.5 − 32 ⎞
c. P( x > 50 ) = P( 51 ≤ x ) = P ⎜ ≤ z ⎟ = P(3.56827 ≤ z ) = 0.0001797. This
⎝ 5.18459 ⎠
probability is very close to zero, so if this outcome occurs, one would be led to doubt the 16%
figure.
⎛ 250 − 266 300 − 266 ⎞

7.124 a. P(250 < x < 300)= P ⎜ <z< ⎟ = P(−1 < z < 2.13) = 0.9834 − 0.1587 =
⎝ 16 16 ⎠
0.8247.
⎛ 240 − 266 ⎞
b. P(x < 240) = P ⎜ z < ⎟ = P(z < −1.63) = 0.0516.
⎝ 16 ⎠
c. P(x is within 16 days of the mean duration)

= P(250 < x < 282) = P(−1 < z < 1) = 0.8413 − 0.1587 = 0.6826.
⎛ 310 − 266 ⎞
d. P(310 ≤ x) = P ⎜ < z ⎟ = P(2.75 < z) = 1 − P(z ≤ 2.75) = 1 − 0.9970 = 0.0030.
⎝ 16 ⎠
The chances of a pregnancy having a duration of at least 310 days is 0.003. This is a small
value, so there is a bit of skepticism concerning the claim of this lady.
e. If the duration is 261 days or less, then the date of birth will be 261 + 14 = 275 days or less
after coverage began. Hence the insurance company will not pay the benefits.
⎛ 261 − 266 ⎞
P(x < 261) = P ⎜ z < ⎟ = P(z < −0.31) = 0.3783 ≈ 38%.
⎝ 16 ⎠
When date of conception is 14 days after coverage began, about 38% of the time the
insurance company will refuse to pay the benefits because of the 275 day requirement.
188
Chapter 8
8.1 A statistic is any quantity computed from the observations in a sample. A population characteristic is
a quantity which describes the population from which the sample was taken.
8.2 The quantity μ denotes the mean of the population and hence it is a population characteristic; x
denotes the sample mean and hence it is a statistic.
The quantity σ denotes the standard deviation of the population; so, it is a population characteristic.
The quantity s denotes the standard deviation of the sample and hence it is a statistic.
8.3 a. population characteristic

b. statistic
c. population characteristic
d. statistic
e. statistic
8 + 14 + 16 + 10 + 11 59
8.4 a. μ= = = 11.8
5 5
b. One possible random sample consists of elements 1 and 4, whose values are 8 and 10. The
value of x for this sample is (8 + 10)/2 = 9.
c. Twenty-four additional samples of size 2 and their means are:
Sample Sample mean Sample Sample mean

14, 16 15.0 10, 16 13.0
16, 14 15.0 10, 11 10.5
11, 14 12.5 11, 14 12.5
11, 10 10.5 16, 8 12.0
14, 10 12.0 16, 8 12.0
16, 11 13.5 8, 10 9.0
14, 10 12.0 11, 10 10.5
8, 11 9.5 8, 10 9.0
16, 14 15.0 10, 8 9.0
16, 10 13.0 8, 10 9.0
10, 11 10.5 16, 14 15.0
8, 14 11.0 10, 16 13.0
d.
0.4
0.3
Density
0.2
0.1
0.0
9.0 9.5 10.0 10.5 11.0 11.5 12.0 12.5 13.0 13.5 14.0 14.5 15.0
Sample Mean
189
For this group of 25 samples of size two, the sample means are either close to the population
mean, or at either extreme. The sample means tend to differ quite a bit from one sample to
the next.
8.5 Ten additional random samples of size 5 and their sample means are:
Sample Sample
mean
323, 261, 159, 222, 342 261.4
168, 184, 323, 230, 261 233.2
159, 295, 323, 261, 275 262.6
262, 231, 263, 222, 168 229.2
275, 261, 270, 159, 319 256.8
159, 261, 323, 258, 267 253.6
262, 168, 261, 333, 275 259.8
267, 222, 295, 231, 261 255.2
168, 230, 267, 184, 262 222.2
231, 323, 159, 261, 265 247.8
The value of x differs from one sample to the next. Samples 1 and 3 produced x values greater
than μ while the remaining eight samples yielded x values less than μ. Samples 2, 4, and 9
produced x values which differ from μ by quite a large amount, while samples 1, 3, 5, 6, 7, 8, and 10
produced x values quite close to μ. These x values are in general agreement with the results
summarized in Figure 8.4 of the text. One area of disagreement may be that 7 of the ten values of x
were less than μ, whereas in Figure 8.4, 22 out of 45 (roughly half) of the x values were less than μ.
8.6 If samples of size ten rather than size five had been used, the histograms would be similar in that
they both would be centered close to 260.25. They would differ in that the histogram based on n = 10
would have less variability than the histogram based on n = 5 (see Figure 8.1 of the text).
8.7 a.
Sample Sample Sample Sample
Mean Mean
1, 2 1.5 3, 1 2.0
1, 3 2.0 3, 2 2.5
1, 4 2.5 3, 4 3.5
2, 1 1.5 4, 1 2.5
2, 3 2.5 4, 2 3.0
2, 4 3.0 4, 3 3.5
The sampling distribution of x based on n = 2, when sampling is done

without replacement, is
Value of x 1.5 2.0 2.5 3.0 3.5

Probability 2/12 2/12 4/12 2/12 2/12
190
Histogram of Sample Mean
0.7
0.6
0.5
0.4
Density 0.3
0.2
0.1
0.0
1.5 2.0 2.5 3.0 3.5
Sample Mean
b.
Sample Sample Mean Sample Sample Mean

1, 1 1.0 3, 1 2.0
1, 2 1.5 3, 2 2.5
1, 3 2.0 3, 3 3.0
1, 4 2.5 3, 4 3.5
2, 1 1.5 4, 1 2.5
2, 2 2.0 4, 2 3.0
2, 3 2.5 4, 3 3.5
2, 4 3.0 4, 4 4.0
The sampling distribution of x based on n = 2, when sampling is done

with replacement, is
Value of x 1.0 1.5 2.0 2.5 3.0 3.5 4.0

Probability 1/16 2/16 3/16 4/16 3/16 2/16 1/16
Histogram of Sample Mean
0.5
0.4
0.3
Density
0.2
0.1
0.0
1.0 1.5 2.0 2.5 3.0 3.5 4.0
Sample Mean
191
c. They are similar in that they are both symmetric about the population mean of 2.5. Also, both
have largest probability at 2.5.
They are different in that, the values in a cover a smaller range, 1.5 to 3.5, than in part b, 1.0
to 4.0. Also, the probability that x will take a value within 0.5 unit of the true mean is 8/12 in
part a whereas it is only 10/16 (a smaller value) in part b.
8.8 We selected a random sample of size n=2 from the population of Exercise 8.7 fifty times and
calculated the sample mean for each sample. The results are shown in the table below. Your own
simulation may produce different results.
Trial First Second Sample Trial First Second Sample

number sample sample mean number sample sample mean
value value value value
1 4 1 2.5 26 3 2 2.5
2 4 3 3.5 27 3 1 2.0
3 2 4 3.0 28 2 3 2.5
4 4 2 3.0 29 2 1 1.5
5 3 1 2.0 30 4 2 3.0
6 4 3 3.5 31 4 1 2.5
7 3 1 2.0 32 2 3 2.5
8 4 2 3.0 33 4 3 3.5
9 2 3 2.5 34 3 2 2.5
10 1 4 2.5 35 4 1 2.5
11 4 3 3.5 36 1 4 2.5
12 1 4 2.5 37 1 4 2.5
13 3 1 2.0 38 1 2 1.5
14 3 4 3.5 39 4 1 2.5
15 2 3 2.5 40 2 4 3.0
16 3 2 2.5 41 4 3 3.5
17 4 2 3.0 42 3 2 2.5
18 2 3 2.5 43 2 4 3.0
19 3 2 2.5 44 1 2 1.5
20 1 2 1.5 45 1 4 2.5
21 4 3 3.5 46 3 4 3.5
22 2 1 1.5 47 3 4 3.5
23 3 2 2.5 48 1 2 1.5
24 4 1 2.5 49 4 1 2.5
25 2 1 1.5 50 4 1 2.5
The relative frequency distribution of the 50 simulated values of x and the corresponding
probabilities from the sampling distribution of x are as follows.
Value of x Relative frequency from Probability from

simulation sampling distribution
1.5 0.14 0.167
2 0.08 0.167
2.5 0.46 0.333
3 0.14 0.167
3.5 0.18 0.167
192
0.9
0.8
0.7
0.6
Density
0.5
0.4
0.3
0.2
0.1
0.0
1.5 2.0 2.5 3.0 3.5
Sample Mean
The relative frequencies from the simulation are approximately equal to the corresponding
probabilities from the sampling distribution of x . We expect the approximations to improve as the
number of simulations increases. Your own simulations may lead to somewhat different results.
8.9 Note: Statistic #3 is sometimes called the midrange.
Sample Value of Mean Value of Median Value of statistic #3

(midrange)
2, 3, 3* 2.67 3 2.5
2, 3, 4 3.00 3 3.0
2, 3, 4* 3.00 3 3.0
2, 3*, 4 3.00 3 3.0
2, 3*, 4* 3.00 3 3.0
2, 4, 4* 3.33 4 3.0
3, 3*, 4 3.33 3 3.5
3, 3*, 4* 3.33 3 3.5
3, 4, 4* 3.67 4 3.5
3*, 4, 4* 3.67 4 3.5
Sampling distribution of statistic #1.
Value of x 2.67 3 3.33 3.67

Probability 0.1 0.4 0.3 0.2
Value of 3 4
median
Probability 0.7 0.3
Value of 2.5 3 3.5

midrange
Probability 0.1 0.5 0.4
Statistic #1 is the only one that is unbiased, but because it has a large standard deviation, there is a
good chance that an estimate would be far away from the value of the population mean. For this
population, Statistic #3 may be a better choice because it is more likely to produce an estimate that is
close to the population mean.
193
10
8.10 a. μ x = 100, σ x = = 3.333
9
10
b. μ x = 100, σ x = = 2.582
15
10
c. μ x = 100, σ x = = 1.667
36
10
d. μ x = 100, σ x = = 1.414
50
10
e. μ x = 100, σ x = = 1.0
100
10
f. μ x = 100, σ x = =0.5
400
8.11 For n = 36, 50, 100, and 400
8.12 σ represents the standard deviation of the population being sampled, while σ x represents the
standard deviation of the sampling distribution of x . Similarly, μ represents the mean value of the
population being sampled, while μ x represents the mean value of the sampling distribution of x .
5 5
8.13 a. μ x = 40, σ x = =
= 0.625
64 8
Since n = 64, which exceeds 30, the shape of the sampling distribution will be approximately
normal.
39.5 − 40 40.5 − 40
b. P( μ − 0.5 < x < μ + 0.5) = P(39.5 < x < 40.5) = P( <z< )
0.625 0.625
= P(−0.8 < z < 0.8) = 0.7881 − 0.2119 = 0.5762
39.3 − 40 40.7 − 40
c. P( x < 39.3 or x > 40.7) = P(z < or z > )
0.625 0.625
= 1 − P(−1.12 < z < 1.12) = 1 − [0.8686 − 0.1314] = 0.2628
0.289
8.14 a. μ x = 0.5, σ x = = 0.072
16
0.289
b. μ x =0.5, σ x = = 0.041
50
We expect the sampling distribution of x to be approximately normal with mean equal to 0.5
and standard deviation equal to 0.041. Thus, the following figure should provide an
approximation for the density curve for the sampling distribution of x .
194
0.8
8.15 a. μ x = 2 and σ x = = 0.267
9
0.8
b. For n = 20, μ x = 2 and σ x = = 0.179
20
0.8
For n = 100, μ x = 2 and σ x = = 0.08
100
In all three cases μ x has the same value, but the standard deviation of x decreases as n
increases. A sample of size 100 would be most likely to result in an x value close to μ. This
is because the sampling distribution of x for n = 100 has less variability than those for n = 9
or 20.
8.16 a. μ x = 150 pounds
27
b. σx = = 6.75 pounds
16
c. The total weight limit of 2500 pounds will be exceeded if the average weight is greater than
2500/16 = 156.25 pounds.
156.25 − 150
d. The required probability is equal to P( x > 156.25) = P( z > ( ) = P ( z > 0.9259 )
6.75
= 1 – P( z ≤ 0.9259 ) = 0.1772.
5
8.17 a. σx = = 1,
25
64 − 65 67 − 65
P(64 ≤ x ≤ 67) = P( ≤z≤ ) = P(−1 ≤ z ≤ 2) = 0.9772 − 0.1587 = 0.8185
1 1
195
68 − 65
P(68 ≤ x ) = P( ≤ z) = P(3 ≤ z) = 1 − P(z < 3) = 1 − 0.9987 = 0.0013
1
b. Because the sample size is large we can use the normal approximation for the sampling
distribution of x by the central limit theorem. We have
5
σx = = 0 .5 ,
100
P(64 ≤ x ≤ 67) = P((64 − 65)/0.5 < z < (67 − 65)/0.5) = P(−2 < z < 4)
= 1 − 0.0228 = 0.9772
P(68 ≤ x ) = P((68 − 65)/.5 ≤ z) = P(6 ≤ z) = 1 − P(z < 6) ≈ 1 − 1 = 0
10
8.18 a. σx = =1
100
x will be within 2 of the value of μ if μ − 2 ≤ x ≤ μ + 2.
( μ − 2) − μ ( μ + 2) − μ
P( μ − 2 ≤ x ≤ μ + 2 ) = P[ ≤z≤ ] = P(−2 ≤ z ≤ 2)
1 1
= 0.9772 − 0.0228 = 0.9544
b. Approximately 95% of the time, x will be within 2 σ x = 2(1) = 2 of μ.
Approximately 0.3% of the time, x will be further than 3 σ x = 3(1) = 3 from μ.
0.02
8.19 If the true process mean is equal to 0.5 in, then μ x = 0.5, σ x = =0.00333
36
P(the line will be shut down unnecessarily) = 1 − P(0.49 ≤ x ≤ 0.51)
0.49 − 0.50 0.51 − 0.50
= 1 − P( ≤z≤ )
0.00333 0.00333
= 1 − P(−3 ≤ z ≤ 3) = 1 − (0.9987 − 0.0013) = 0.0026
8.20 a.
0.04
0.03
Density
0.02
0.01
0.00
0 5 10 15 20 25 30 35 40 45
Number of checks written
The distribution shown above is bimodal with the mode around 12 checks being associated
with the population of college students and the mode around 32 being associated with
nonstudent residents.
196
16.5
b. μ x = 22.0, σ x = = 1.65
100
c. The central limit theorem is the justification for the following computations.
20 − 22
P( x ≤ 20) = P(z ≤ ) = P(z ≤ −1.2121) = 0.1127
1.65
25 − 22
P( x ≥ 25) = P(z ≥ ) = P(z ≥1.82) = 1 − P(z < 1.82) = 1 − 0.9655 = 0.0345
1.65
8.21 The total weight of their baggage will exceed the limit, if the average weight exceeds 6000/100 = 60.
20
μ x = 50, σ x = = 2,
100
60 − 50
P(total weight ≥ 6000) = P( x ≥ 60) = P(z ≥ ) = P(z ≥ 5) = 1 − P(z < 5) ≈ 1 − 1 = 0
2
8.22 a. Because the x distribution is normal, the x distribution will be normal (even for n less than
30).
0.05
μ x = 3; σ x = = 0.0125
16
b. 3 + 3 σ x = 3 + 3(0.0125) = 3.0375 3 − 3 σ x = 3 − 3(0.0125) = 2.9625
c. P( x would be outside 3 ± 3 σ x ) = 1− P( 3 − 3 σ x ≤ x ≤ 3 + 3 σ x ) = 1− P(− 3 ≤ z ≤ 3)

= 1−(0.9987− 0.0013) = 0.0026
d. P(the problem is detected) = P( x < 2.9625 or x > 3.0375)

2.9625 − 3.05 3.0375 − 3.05
= P(z < or z > )
0.0125 0.0125
= P(z < −7 or z > −1) = P(z < −7) + P(z > −1)
= 0 + (1 − 0.1587) = 0.8413
197
8.23 a. μ p = 0.65, σ p = 0.65(0.35) /10 = 0.15083
b. μ p = 0.65, σ p = 0.65(0.35) / 20 = 0.10665
c. μ p = 0.65, σ p = 0.65(0.35) / 30 = 0.08708
d. μ p = 0.65, σ p = 0.65(0.35) / 50 = 0.06745
e. μ p =0.65, σ p = 0.65(0.35) /100 = 0.04770
f. μ p = 0.65, σ p = 0.65(0.35) / 200 = 0.03373
8.24 When π = 0.65;

n = 10, nπ = 10(0.65) = 6.5, n(1 − π) = 10(0.35) = 3.5; neither are ≥ 10
n = 20, nπ = 20(0.65) = 13, n(1 − π) = 20(0.35) = 7; only nπ ≥ 10
n = 30, nπ = 30(0.65) = 19.5, n(1 − π) = 30(0.35) = 10.5; both are ≥ 10
So, n = 30, 50, 100, and 200.
When π = 0.2;
n = 10, nπ = 10(0.2) = 2, n(1 − π) = 10(0.8) = 8 : neither are ≥ 10
n = 20, nπ = 20(0.2) = 4, n(1 − π ) = 20(0.8) = 16 : only n(1 − π ) ≥ 10
n = 30, nπ = 30(0.2) = 6, n(1 − π) = 30(0.8) = 24 : only n(1 − π ) ≥ 10
n = 50, nπ = 50(0.2) = 10, n(1 − π) = 50(0.8) = 40 : both ≥ 10
So n = 50, 100, and 200.
(0.07)(0.93)
8.25 a. μ p = 0.07, σ p = = 0.0255
100
b. No; nπ = 100(0.07) = 7, and n(1 − π) = 100(1 - 0.07) = 93. For the sampling distribution of p
to be considered approximately normal, both have to be greater or equal to 10.
c. The value of the mean doesn’t change as it isn’t dependent on the sample size. The
(0.07)(0.93)
standard deviation becomes : σ p = = 0.01804
200
d. Yes; nπ = 200(0.07) = 14, and n(1 − π) = 200(1 - 0.07) = 186. For the sampling distribution of
p to be considered approximately normal, both have to be greater or equal to 10.
0.1 − 0.07
e. P(p>0.1) = P(z > ) = P(z > 1.66) = 0.0485
0.01804
(0.15)(0.85)
8.26 a. μ p = 0.15, σ p = = 0.0357
100
b. Yes; nπ = 100(0.15) = 15, and n(1 − π) = 100(1 - 0.15) = 85. For the sampling distribution of
198
c. The value of the mean doesn’t change as it isn’t dependent on the sample size. The
(0.15)(0.85)
standard deviation becomes: σ p = = 0.02525
200
d. Yes; nπ = 200(0.15) = 30, and n(1 − π) = 200(1 - 0.15) = 170. For the sampling distribution of
0.1 − 0.15
e. P(p>0.1) = P(z > ) = P(z > -1.98) = 0.9761
0.02525
(0.005)(0.995)
8.27 a. μ p = 0.005, σ p = = 0.007
100
b. Since nπ = 100(0.005) = 0.5 is less than 10, the sampling distribution of p cannot be
approximated well by a normal curve.
c. The requirement is that nπ ≥ 10, which means that n would have to be at least 2000.
8.28 a. No, because nπ = 10(0.3) = 3 which does not exceed 10.
0.3(0.7)
b. μ p = 0.3, σ p = = 0.000525 =0.0229
400
c. Since (400)(0.3) and (400)(1 - 0.3) are both greater than 10, a normal approximation for the
sampling distribution of p is reasonable. Hence P(0.25 ≤ p ≤ 0.35)
⎛ 0.25− 0.30 p − 0.3 0.35− 0.30 ⎞
≈ P⎜ ≤ ≤ ⎟ = P(−2.18 < z < 2.18) = 0.9854 − 0.0145 = 0.9709
⎝ 0.0229 0.0229 0.0229 ⎠
d. It is smaller because as n increases, σ P decreases. This means that the z-scores for 0.25
and 0.35 would be larger in absolute value. In fact, when n = 500, σ P = 0.0205, and the z-
scores would be ± 2.44. The area under a normal curve between −2.18 and 2.18 is smaller
than the area between −2.44 and 2.44.
0.5(0.5)
8.29 a. For π = 0.5, μ p = 0.5 and σ p = = 0.0333
225
0.6(0.4)
For π = 0.6, μ p =0.6 and σ p = = 0.0327
225
For both cases, nπ ≥ 10 and n(1 − π) ≥ 10. Hence, in each instance, p would have an
approximately normal distribution.
0.6 − 0.5
b. For π = 0.5, P(p ≥ 0.6) = P(z ≥ ) = P(z ≥ 3) = 1 −P(z < 3)
0.0333
= 1 − 0.9987 = 0.0013.
0.6 − 0.6
For π = 0.6, P(p ≥ 0.6) = P(z ≥ ) = P(z ≥ 0) = 1 − P(z < 0)
0.0327
= 1 − 0.5000 = 0.5000
c. When π = 0.5, the P(p ≥ 0.6) would decrease.

When π = 0.6, the P(p ≥ 0.6) would remain the same.
199
(0.48)(0.52)
8.30 μ p = π = 0.48, σ p = = 0.02234
500
0.5 − 0.48
P(p >0.5) = P(z > ) = P(z > 0.90 ) = 1 − P(z ≤ 0.8953) ≈ 1 − 0.8147 = 0.1853
0.02234
(0.05)(0.95)
8.31 a. μ p = π =0.05, σ p = =0.01541
200
0.02 − 0.05
P(p > 0.02 ) = P(z > ) = P(z > −1.95)
0.01541
= 1 − P(z ≤ −1.95) ≈ 1 − 0.0258 = 0.9742
(0.1)(0.9)
b. μ p = π = 0.10, σ p = = 0.02121
200
0.02 − 0.10
P(p ≤ 0.02) = P(z ≤ ) = P(z ≤ −3.77) ≈ 0
0.02121
0.1
8.32 μ X = 0.8 and σ X = =0.01
100
0.79 − 0.8
P( x < 0.79) = P(z < ) = P(z < −1) ≈ .1587
0.01
0.77 − 0.80
P( x < 0.77) = P(z < ) = P(z < −3) ≈ .0013
0.01
8.33 a. The sampling distribution of x will be approximately normal, with mean equal to 50 (lb) and
1
standard deviation equal to = 0.1 (lb).
100
49.75 − 50 50.25 − 50
b. P(49.75 < x < 50.25) = P( <z< ) = P(−2.5 < z < 2.5)
0.1 0.1
= .9938 − .0062 = .9876
50 − 50
c. P( x < 50) = P(z < ) = P(z < 0) = 0.5
0.1
(0.2)(0.8)
8.34 π = 0.20, n = 100, μ p =0.2, σ p = =0.04
100
0.25 − 0.20
P(p > 0.25) = P(z > ) = P(z > 1.25) = 1 − P(z ≤ 1.25) = 1 − 0.8944 = 0.1056
0.04
200
850 − 1000 1300 − 1000
8.35 a. P(850 < x < 1300) = P( <z< ) = P(−1 < z < 2)
150 150
= 0.9772 − 0.1587 = 0.8185
150
b. μ x = 1000, σ x = = 47.43
10
950 − 1000 1100 − 1000

P(950 < x < 1100) = P( <z< )
47.43 47.43
= P(−1.054 < z < 2.108) = 0.9825 − 0.1459 = 0.8366
850 − 1000 1300 − 1000

P(850 < x < 1300) = P( <z< )
47.43 47.43
= P(−3.16 < z < 6.33) = 1 − 0.0008 = 0.9992
(0.4)(0.6)
8.36 μ p =0.4, σ p = =0.049
100
0.5 − 0.4 p − 0.4

a. P(0.5 ≤ p) ≈ P( ≤ ) = P(2.04 ≤ z) = 1 − P(z < 2.04)
0.049 0.049
= 1 − 0.9793 = 0.0207
b. The probability of obtaining a sample of 100 in which at least 60 participated in such a plan is
almost 0 if π = 0.40. Hence, one would doubt that π = 0.40 and strongly suspect that π has a
value greater than 0.4.
30
8.37 μ = 100, σ = 30, μ x = 100, σ x = = 4.2426
50
106 − 100
P( total> 5300) = P( x > 106) = P(z > ) = P(z > 1.4142 < z) = 1 − P(z ≤ 1.4142)
4.2426
= 1 − 0.9214 = 0.0786
201
Chapter 9
9.1 Statistic II would be preferred because it is unbiased and has smaller variance than the other two.
9.2 An unbiased statistic is generally preferred over a biased statistic, because there is no long run
tendency for the unbiased statistic to overestimate or underestimate the true population value. That
is, there will be no systematic estimation error.
Unbiasedness by itself does not guarantee that the estimate will be close to the true value. An
unbiased statistic may have a sampling distribution that has a large variance. Thus, it would be
possible to obtain a value of the statistic that is quite inaccurate. One might choose a biased statistic
over an unbiased one if the bias of the first is small and if its sampling distribution has a small
variance.
1720
9.3 p= = 0.2769
6212
9.4 a. x=
∑ x = 27678 = 1845.2
n 15
b. x=
∑ x = 26626 = 1775.1
n 15
c. No fast food is consumed: s = 386.346, fast food is consumed: s = 620.660
number in sample registered 14

9.5 The point estimate of π would be p= = = 0.70
n 20
245
9.6 p = 935 = 0.262
19.57
9.7 a. = 1.957
10
b. s2 = 0.15945
c. s = 0.3993; No, this estimate is not unbiased. It underestimates the true value of σ.
392.4
9.8 a. x= = 20.6526
19
b. The number of cyclists in the sample whose gross efficiency is at most 20 is 4. The estimate
of all such cyclists whose gross efficiency is at most 20 is the sample proportion p = 4/19 =
0.2105.
9.9 a. The value of σ will be estimated by using the statistic s. For this sample,
∑ x 2 = 1757.54, ∑ x = 143.6, n = 12
( ∑ x )2 (143.6)2
∑ x2 − 1757.54 −
2 n = 12 1757.54 − 1718.4133
s = =
n− 1 12 − 1 11
39.1267
= = 3.557 and s = 3.557 = 1.886
11
203
b. The population median will be estimated by the sample median. Since n = 12 is even, the
sample median equals the average of the middle two values (6th and 7th values), i.e.,
(11.3 + 11.4) = 11.35.
2
c. In this instance, a trimmed mean will be used. First arrange the data in increasing order.
Then, trimming one observation from each end will yield an 8.3% trimmed mean. The
trimmed mean equals 117.3/10 = 11.73.
d. The point estimate of μ would be x = 11.967. From part a, s = 1.886. Therefore the
estimate of the 90th percentile is 11.967 + 1.28(1.886) = 14.381.
9.10 a. x J = 120.6
b. An estimate of the total amount of gas used by all these houses in January would be
10000(120.6) = 1,206,000 therms. More generally, an estimate of the population total is
obtained by multiplying the sample mean by the size of the population.
c. p = 8/10 = 0.8
d. Using the sample median, an estimate of the population median usage is

(118 + 122)/2 = 120 therms.
9.11 a. As the confidence level increases, the width of the large sample confidence interval also
increases.
b. As the sample size increases, the width of the large sample confidence interval decreases.
9.12 The values for parts a - d are found in the table for Standard Normal Probabilities (Appendix Table 2).
a. 1.96
b. 1.645
c. 2.58
d. 1.28
e. 1.44 (approximately)
9.13 For the interval to be appropriate, np ≥ 10, n(1 − p) ≥ 10 must be satisfied.

a. np = 50(0.3) = 15, n(1 − p) = 50(0.7) = 35, yes
b. np = 50(0.05) = 2.5, no
c. np = 15(0.45) = 6.75, no
d. np = 100(0.01) = 1 , no
e. np = 100(0.70) = 70, n(1 − p) = 100(0.3) = 30, yes
f. np = 40(0.25) = 10, n(1 − p) = 40(0.75) = 30, yes
g. np = 60(0.25) = 15, n(1 − p) = 60(0.75) = 45, yes
h. np = 80(0.10) = 8, no
204
9.14 a. As the confidence level increases, the width of the confidence interval for π increases.
b. As the sample size increases, the width of the confidence interval for π decreases.
c. As the value of p gets farther from 0.5, either larger or smaller, the width of the confidence
interval for π decreases.
9.15 a. Because np = 420.42 and n(1 − p ) = 580.58 , which are both greater than 10, and the
Americans in the sample were randomly selected from a large population, the large-sample
interval can be used. The sample proportion p is 0.42. The 99% confidence interval for π ,
the proportion of all Americans who made plans in May 2005 based on an incorrect weather
0.42(1 − 0.42)
report would be 0.42 ± 2.58 ⇒ 0.42 ± 0.0402 ⇒ (0.3798, 0.4602). We can
1001
be 99% confident that the true proportion of adult Americans who made plans in May 2005
based on an incorrect weather report is between .38 and .46.
b. No, weather reports may be more or less reliable during other months.
9.16 a. Because np = 370 and n(1 − p ) = 630 , which are both greater than 10, and the students in
the sample were randomly selected from a large population, the large-sample interval can be
used. The sample proportion p is 0.37. The 90% confidence interval for π , the proportion of
all college freshman who carried a credit card balance from month to month would be
0.37(1 − 0.37)
0.37 ± 1.645 ⇒ 0.37 ± 0.02511 ⇒ (0.3449, 0.3951).
1000
b. Because np = 480 and n(1 − p ) = 520 , which are both greater than 10, and the students in
the sample were randomly selected from a large population, the large-sample interval can be
used. The sample proportion p is 0.48. The 90% confidence interval for π , the proportion of
all college seniors who carried a credit card balance from month to month would be
0.48(1 − 0.48)
0.48 ± 1.645 ⇒ 0.48 ± 0.02599 ⇒ (0.45401, 0.50599).
1000
c. The estimated standard deviation of the sampling distribution of p is larger when p is 0.48
than when p = 0.37 so the width of the confidence interval will be wider.
9.17 a. Because np and n(1 − p ) are both greater than 10, and the potential jurors in the sample
were randomly selected from a large population, the large-sample interval can be used. The
sample proportion p is 350/500 = 0.7. The 95% confidence interval for π , the population
0.7(1 − 0.7)
proportion would be 0.7 ± 1.96 ⇒ 0.7 ± 0.0402 ⇒ (0.6598, 0.7402). With 95%
500
confidence we can estimate that between 66% and 74% of all potential jurors regularly watch
at least one crime-scene investigation series.
b. A 99% confidence interval would be wider than the 95% confidence interval in Part (a).
9.18 a. Because np and n(1 − p ) are both greater than 10, and the adults in the sample were
randomly selected from a large population, the large-sample interval can be used. The
sample proportion p is 230/1000 = 0.23. The 95% confidence interval for π , the proportion
of all U.S. adults for whom math was the favorite subject would be
0.23(1 − 0.23)
0.23 ± 1.96 ⇒ 0.23 ± 0.0261 ⇒ (0.2039, 0.2561).
1000
205
b. Because np and n(1 − p ) are both greater than 10, and the adults in the sample were
of all U.S. adults for whom math was the least favorite subject would be
0.37(1 − 0.37)
0.37 ± 1.96 ⇒ 0.37 ± 0.0299 ⇒ (0.3401, 0.3999).
1000
9.19 a. Because np and n(1 − p ) are both greater than 10, and the businesses in the sample were
sample proportion p is 137/526 = 0.26. The 95% confidence interval would be
0.26(1 − 0.26)
0.26 ± 1.96 ⇒ 0.26 ± 0.0375 ⇒ (0.2225, 0.2975). With 95% confidence we
526
can estimate that between 22.25% and 29.75% of all U.S. businesses have fired workers for
misuse of the Internet.
b. It would be narrower because of a lower confidence level (90% instead of 95%) and because
of a smaller estimated standard error (0.0189 instead of 0.0191).
of all U.S. adults that consider themselves to be baseball fans would be:
0.394(1 − 0.394)
0.394 ± 1.96 ⇒ 0.394 ± 0.0303 ⇒ (0.3637, 0.4243).
1000
sample proportion p is 272/394 = 0.69. The 95% confidence interval for π , the proportion of
all U.S. adults who consider themselves to be baseball fans that think the designated hitter
should either be expanded to both leagues or eliminated would be:
0.69(1 − 0.69)
0.69 ± 1.96 ⇒ 0.69 ± 0.0457 ⇒ (0.6443, 0.7357).
394
c. The intervals are not the same width (Part b is wider) because the sample size is different
(larger in Part a) and the standard error is different (larger in Part b).
p(1 − p )
9.21 Bound of error (based on 95% confidence) = 1.96 . If p = 0.82 and n = 1002,
n
0.82(1 − 0.82)
1.96 = 0.0238 . This implies a 95% confidence interval of 0.82 ± 0.0238 or (0.7962,
1002
0.8438). With 95% confidence, we can estimate that between 79.6% and 84.4% of all adults think
that reality shows are ‘totally made up’ or ‘mostly distorted’.
sample proportion p is 0.52. The 90% confidence interval for π , the proportion of all
0.52(1 − 0.52)
American adults who think lying is never justified would be: 0.52 ± 1.645 ⇒
1000
0.52 ± .026 ⇒ (0.494, 0.546).
206
of all American adults who think lying is OK if it avoids hurting someone’s feelings would be:
0.65(1 − 0.65)
0.65 ± 1.645 ⇒ 0.65 ± 0.0248 ⇒ (0.6252, 0.6748).
1000
c. It is estimated that about half of all adult Americans think lying is never justified, yet more
than 60% of them think that it is OK to lie if it avoids hurting someone’s feelings. Obviously
there are some adults who consider “lying to avoid hurting someone’s feelings” as not really a
lie!
9.23 a. The sample proportion p is 38/115 = 0.330. The 95% confidence interval would be
0.330(1 − 0.330)
0.330 ± 1.96 ⇒ 0.330 ± 0.0859 ⇒ (0.2441, 0.4159).
115
b. The sample proportion p is 22/115 = 0.1913. The 90% confidence interval would be
0.1913(1 − 0.1913)
0.1913 ± 1.645 ⇒ 0.1913 ± 0.0603 ⇒ (0.1310, 0.2516).
115
c. The interval is wider in part a because (i) the confidence interval is higher in part (a) and (ii)
the sample proportion is more extreme; further from 0.5 in part (a) .
9.24 a. The sample proportion who can identify their own country is 0.9. The 90% confidence interval
0.9(1 − 0.9)
would be 0.9 ± 1.645 ⇒ 0.9 ± 0.00901 ⇒ (0.891, 0.909).
3000
b. The sample should be a SRS of the respondents and independent of each other.
c. The results would only apply to the population of respondents aged 18 to 24 in the nine
different countries chosen for the study.
9.25 The sample proportion p is 18/52 = 0.3462. The 95% confidence interval would be
0.3462(1 − 0.3462)
0.3462 ± 1.96 ⇒ 0.3462 ± 0.1293 ⇒ (.2169, .4755).
52
The assumption are: p is the sample proportion from a random sample, and that the sample size is
large, (np ≥ 10, and n(1 - p) ≥ 10).
(.36)(.64)
9.26 With p = .36 and n = 1004, the bound on error is 1.96 ≈ .03
1004
(.25)(.75)
9.27 With p = .25 and n = 1002, the bound on error is 1.96 ≈ .03
1002
9.28 An estimate of the proportion of school children in Australia who watch TV before school is p =
1060/1710 = 0.6199. A 95% confidence interval for the true proportion is
0.6199(1 − 0.6199)
0.6199 ± 1.96 ⇒ 0.6199 ± 0.023 ⇒ (0.597, 0.643) .
1710
For this confidence interval to be valid, the sample must be a random sample from the population of
interest.
2 2
⎡1.96 ⎤ ⎡ 1.96 ⎤
9.29 n = 0.25 ⎢ ⎥ = 0.25 ⎢ 0.05 ⎥ = 384.16; take n = 385.
⎣ B ⎦ ⎣ ⎦
207
9.30 a. 90%
b. 95%
c. 95%
d. 99%
e. 1%
f. 0.5%
g. 5%
9.31 a. 2.12
b. 1.80
c. 2.81
d. 1.71
e. 1.78
f. 2.26
114.4 + 115.6
9.32 a. x is the midpoint of the interval. So, x = = 115.0 .
2
b. As the confidence level increases the width of the interval increases.
Hence (114.4, 115.6) is the 90% interval and (114.1, 115.9) is the 99% interval.
9.33 As the sample size increases, the width of the interval decreases. The interval (51.3, 52.7) has a
width of 52.7 − 51.3 = 1.4 and the interval (49.4, 50.6) has a width of 50.6 − 49.4 = 1.2. Hence, the
interval (49.4, 50.6) is based on the larger sample size.
9.34 a. The 90% confidence interval would have been narrower, since its z critical value would have
been smaller.
b. The statement is incorrect. The 95% refers to the percentage of all possible samples that
result in an interval that includes μ, not to the chance (probability) that a specific interval
contains μ.
c. Again this statement is incorrect. While we would expect approximately 95 of the 100
intervals constructed to contain μ, we cannot be certain that exactly 95 out of 100 of them
will. The 95% refers to the percentage of all possible intervals that include μ.
9.35 Because the specimens were randomly selected and the distribution of the breaking force is
approximately normal, the t confidence interval formula for the mean can be used. The 95%
confidence interval for μ , the population mean breaking force is:
⎛ 41.97 ⎞
306.09 ± 2.57 ⎜ ⎟ ⇒ 306.09 ± 44.035 ⇒ ( 262.06, 350.125 ) . With 95% confidence we can
⎝ 6 ⎠
estimate the average breaking force for acrylic bone cement to be between 272.5 and 229.7
Newtons.
9.36 a. The two groups ‘12 to 23 months’ and ‘24 to 35 months’ would both have the same variability
as each other but greater variability than the ‘less than 12 months group’ because the interval
width for both groups is 0.4 which is wider than the ‘less than 12 month’ group with an interval
width of 0.2.
b. The group ‘less than 12 months’ would have the largest sample size because the interval
width for that group is narrower that for the other groups.
c. It would be a 99% confidence level because if everything else remains constant, an increase
in the confidence level results in an increase in the interval width. The new confidence
interval is wider.
208
9.37 Because the adults were randomly selected and the sample size is large, the t confidence interval
formula for the mean can be used. The 90% confidence interval for μ , the population mean
24.2
commute time is: 28.5 ± 1.645 ⇒ 28.5 ± 1.780 ⇒ ( 26.72, 30.28 ) . With 90% confidence, we
500
estimate that the mean daily commute time for all working residents of Calgary, Canada is between
26.7 minutes and 30.3 minutes.
9.38 a. If the distribution of the anticipated Halloween expense is heavily skewed to the right, the
standard deviation could be greater than the mean.
b. The distribution of anticipated Halloween expense is not approximately normal. As expense

has to be a nonnegative value, and the standard deviation is larger than the mean, the
distribution of Halloween expense is heavily skewed to the right.
c. Even though the distribution of expense is not approximately normal the distribution for the
sampling distribution when n = 1000 will be approximately normal and we can use the t
confidence interval to estimate the mean anticipated Halloween expense for Canadian
residents.
d. Because the adults were randomly selected and the sample size is large, the t confidence
interval formula for the mean can be used. The 99% confidence interval for μ , the
population mean anticipated expense is:
83.7
46.65 ± 2.58 ⇒ 46.65 ± 6.829 ⇒ ( 39.82, 53.48 )
1000
With 99% confidence, we estimate the mean anticipated Halloween expense for all Canadian
residents to be between $39.82 and $53.48.
9.39 a. The t critical value for a 95% confidence interval when df = 99 is 1.99. The confidence
interval based on this sample data is
s ⎛ 20 ⎞
x ± (t critical) ⇒ 183 ± (1.99) ⎜ ⎟ ⇒ (179.03, 186.97) .
n ⎝ 100 ⎠
s ⎛ 23 ⎞
b. x ± (t critical) ⇒ 190 ± (1.99) ⎜ ⎟ ⇒ (185.44, 194.56)
n ⎝ 100 ⎠
c. The new FAA recommendations are above the upper level of both confidence levels so it
appears that Frontier airlines have nothing to worry about.
range 20.3 − 19.9

9.40 A reasonable estimate of σ is = = 0.1 . The required sample size
4 4
⎡ (1.96 )( 0.1) ⎤
2 2
⎡ 1.96σ ⎤
n= ⎢ ⎥ =⎢ ⎥ = 3.84 . A sample size of 4 or large would be required.
⎣ B ⎦ ⎣ 0.1 ⎦
9.41 a. The t critical value for a 90% confidence interval when df = 9 is 1.83. The confidence interval
based on this sample data is
s ⎛ 3.6757 ⎞
x ± (t critical) ⇒ 54.2 ± (1.83) ⎜ ⎟ ⇒ 54.2 ± 2.1271 ⇒ (52.073, 56.327) .
n ⎝ 10 ⎠
b. If the same sampling method was used to obtain other samples of the same size and
confidence intervals were calculated from these samples, 90% of them would contain the true
population mean.
209
c. As airlines are often rated by how often their flights are late, I would recommend the
published arrival time to be close to the upper bound of the confidence interval of the journey
time: 10: 57 a.m.
s
9.42 The standard error is , and so is dependent on the sample size. Consider the following table:
n
Standard Standard
Error deviation
Hispanic s s $77001.58
= 3011
n 654
Native American s s $106641.39
= 29577
n 13
Now the variability doesn’t seem quite as different.
9.43 A boxplot (below) shows the distribution to be slightly skewed with no outliers. It seems plausible
that the population distribution is approximately normal. Calculation of a confidence interval for the
population mean cadence requires sample mean and sample standard deviation:
x = 0.926 s = 0.0809 t critical value with 19 df is 2.58.
⎛ 0.0809 ⎞
The interval is 0.926 ± (2.86) ⎜ ⎟ = 0.926 ± 0.052 = (0.874,0.978) . With 99% confidence, we
⎝ 20 ⎠
estimate the mean cadence of all healthy men to be between 0.874 and 0.978 strides per second.
0.80 0.85 0.90 0.95 1.00 1.05

Cadence
9.44 The t critical value for a 90% confidence interval when df = 10 − 1 = 9 is 1.83. From the given data, n
219
= 10, ∑ x = 219, and ∑ x 2 = 4949.92 . From the summary statistics, x = = 21.9
10
(219)2
4949.92 −
2 10 = 4949.92 − 4796.1 = 153.82 = 17.09
s =
9 9 9
s = 17.09 = 4.134 .
The 90% confidence interval based on this sample data is
s 4.134
x ± (t critical ) ⇒ 21.9 ± (1.83) ⇒ 21.9 ± 2.39 ⇒ (19.51, 24.29 ) .
n 10
210
9.45 Summary statistics for the sample are: n = 5, x = 17, s = 9.03
The 95% confidence interval is given by
s 9.03
x ± (t critical) ⇒ 17 ± (2.78) ⇒ 17 ± 11.23 ⇒ (5.77, 28.23) .
n 5
9.46 a. Summary statistics for the sample are: n = 900, x = 3.73, s = 0.45
The 95% confidence interval is given by
s 0.45
x ± (t critical) ⇒ 3.73 ± (1.96) ⇒ 3.73 ± 0.0294 ⇒ (3.70, 3.76) .
n 900
b. No. With 95% confidence we can state that the mean GPA for the students at this university
is between 3.70 and 3.76. The individual GPAs of the students will vary quite a bit more.
Since approximately 95% of the population values will lie between μ − 2σ and μ + 2σ by the
empirical rule, we estimate that approximately 95% of the students will have their GPAs in the
interval x − 2s and x + 2s , i.e., in the interval 3.73 ± 2(0.45) ⇒ (2.83, 4.63) ; since the maximum
possible GPA is 4.0, we can say that roughly 95% of the Caucasian students will have their
GPAs in the interval from 2.83 to 4.0.
9.47 Since the sample size is small (n = 17), it would be reasonable to use the t confidence interval only if
the population distribution is normal (at least approximately). A histogram of the sample data (see
figure below) suggests that the normality assumption is not reasonable for these data. In particular,
the values 270 and 290 are much larger than the rest of the data and the distribution is skewed to the
right. Under the circumstances the use of the t confidence interval for this problem is not reasonable.
4
Frequency
120 140 160 180 200 220 240 260 280

Calories per half cup
9.48 An estimate σ is (700 − 50)/4 = 650/4 = 162.5. The required sample size is
⎡1.96 (162.5 ) ⎤
2
⎥ = ( 31.85 ) = 1014.42. So take n = 1015.

2
n=⎢
⎣ 10 ⎦
211
⎡ ( z critical ) σ ⎤ ⎡ (1.96 )(1) ⎤
2 2
⎥ = (19.6 ) = 384.16 . Hence, n should be 385.

2
9.49 n=⎢ ⎥ =⎢
⎣ B ⎦ ⎣ 0.1 ⎦
⎡ (1.645 ) σ ⎤
2
9.50 For 90% confidence level: n= ⎢ ⎥

⎣ B ⎦
⎡ ( 2.33 ) σ ⎤
2
For 98% confidence level: n = ⎢ ⎥

⎣ B ⎦
466
9.51 p= = 0.4596
1014
2
⎡ 1.96 ⎤
9.52 The required sample size is n = (.27)(.73) ⎢ ⎥ = 302.87 ; at least 303
⎣ 0.05 ⎦
2
⎡ 1.96 ⎤
The required sample size is n = (.5)(.5) ⎢ ⎥ = 384.16 ; at least 385
⎣ 0.05 ⎦
I would recommend the larger sample size; if I was going to the trouble of surveying 303 adult
residents in my city, it wouldn’t take much more effort to find 82 more!
0.65(1 − 0.65)
9.53 A 90% confidence interval is 0.65 ± 1.645 ⇒ 0.65 ± 0.064 ⇒ (0.589, 0.714)
150
Thus, we can be 90% confident that between 58.9% and 71.4% of Utah residents favor fluoridation.
This is consistent with the statement that a clear majority of Utah residents favor fluoridation.
0.4
9.54 a. 0.5 ± 1.96 ⇒ 0.5 ± 0.089 ⇒ ( 0.411, 0.589 )
77
b. The fact that 0 is not contained in the confidence interval does not imply that all students lie
to their mothers. There may be students in the population and even in this sample of 77 who
did not lie to their mothers. Even though the mean may not be zero, some of the individual
data values may be zero. However, if the mean is nonzero, it does imply that some students
tell lies to their mothers.
⎛ 14.41 ⎞
9.55 a. 25.62 ± 2.33 ⎜ ⎟ ⇒ 25.62 ± 5.062 ⇒ ( 20.558, 30.682 )
⎝ 44 ⎠
⎛ 15.31 ⎞
b. 18.10 ± 2.33 ⎜ ⎟ ⇒ 18.10 ± 2.225 ⇒ (15.875, 20.325 )
⎝ 257 ⎠
c. It is based on a larger sample.
d. Since the interval in a gives the plausible values for μ and the lower endpoint is greater than
20, this suggests that the mean number of hours worked per week for non-persistors is
greater than 20.
212
9.56 Let π = the true proportion of full-time workers that felt like hitting a co-worker during the past year.
125
For this sample p = = 0.167 . Because np = 125 and n(1 − p ) = 625 are both greater than or
750
equal to 10, the sample size is large enough to use the formula for a large-sample confidence
interval. The 90% confidence interval is:
0.167 (1 − 0.167 )
0.167 ± 1.645 ⇒ 0.167 ± 0.0224 ⇒ ( 0.1446, 0.1894 ) .
750
Based on these sample data, we can be 90% confident that the true proportion of full time workers
who have been angry enough in the last year to consider hitting a coworker is between .144 and .190.
9.57 Using a conservative value of π = .5 in the formula for required sample size gives:
2 2
⎛ 1.96 ⎞ ⎛ 1.96 ⎞
n = π (1 − π ) ⎜ ⎟ = .25 ⎜ ⎟ = 384.16 . A sample size of at least 385 should be used.
⎝ B ⎠ ⎝ .05 ⎠
0.77(0.23)
9.58 The 95% confidence interval would be 0.77 ±1.96 ⇒ 0.77 ± 0.029 ⇒ (0.741, 0.799) .
800
2 2
⎡ 2.576 ⎤ ⎡ 2.576 ⎤
9.59 n = 0.25 ⎢ ⎥ = 0.25 ⎢ 0.10 ⎥ = 165.9; take n = 166.
⎣ B ⎦ ⎣ ⎦
9.60 Based on the information given, a 95% confidence interval for π , the proportion of the population
who felt that their financial situation had improved during the last year, is calculated as follows.
0.43(1 − 0.43)
0.43 ± 1.96 ⇒ 0.43 ± 0.0318 ⇒ (0.398, 0.462) .
930
Hence, with 95% confidence, the percentage of people who felt their financial situation had improved
during the last year is between 39.8% and 46.2%. These end points of the confidence interval are
3.2% away on either side of the estimated value of 43%. In the report, the value 3.2% has been
rounded down to 3%. Thus, the claim of a 3% “margin of error” in the report is statistically justified.
An alternative way to justify the statement in the article is as follows. The mean of the sampling
π (1 − π )
distribution of p is π and the standard deviation is equal to . The largest possible standard
930
deviation occurs when π is equal to 0.5, in which case the standard deviation is equal to 0.0164.
Hence the probability that p will be within 0.03 (within 3 percent) of π is greater than or equal to P( (-
0.03/0.0164) < z < (0.03/0.0164) ) = P(-1.8297 < z < 1.8297) = 0.933 which is approximately equal to
0.95. Hence the probability is 93.3% or greater that the value of p is within 3 percent of the true value
of π .
9.61 a. The 95% confidence interval for μ ABC is

⎛ 5 ⎞
15.6 ± (1.96) ⎜ ⎟ ⇒ 15.6 ± 1.39 ⇒ (14.21, 16.99) .
⎝ 50 ⎠
b. For μ CBS : 11.9 ± 1.39 ⇒ (10.51, 13.29)
For μ FOX : 11.7 ± 1.39 ⇒ (10.31, 13.09)
For μ NBC : 11.0 ± 1.39 ⇒ (9.61, 12.39)
213
c. Yes, because the plausible values for μ ABC are larger than the plausible values for the other
means. That is, μ ABC is plausibly at least 14.21, while the other means are plausibly no
greater than 13.29, 13.09, and 12.39.
0.721( 0.279 )
9.62 a. 0.721 ± 2.58 ⇒ 0.721 ± 0.052 ⇒ ( 0.669, 0.773 )
500
0.279 ( 0.721)
b. 0.279 ± 1.96 ⇒ 0.279 ± 0.039 ⇒ ( 0.24, 0.318 )
500
Based on this interval, we conclude that between 24% and 31.8% of college freshman are
not attending their first choice of college.
c. It would have been narrower.
9.63 Since n = 18, the degrees of freedom is n − 1 = 18 − 1 = 17 .
From the t-table, the t critical value is 2.11.
⎛ 4.3 ⎞
The confidence interval is 9.7 ± 2.11⎜ ⎟ ⇒ 9.7 ± 2.14 ⇒ ( 7.56, 11.84 ) .
⎝ 18 ⎠
Based on this interval we can conclude with 95% confidence that the true average rating of
acceptable load is between 7.56 kg and 11.84 kg.
101
9.64 p= = 0.0996 . The 90% confidence interval is
1014
0.0996 ( 0.9004 )
0.0996 ± 1.645 ⇒ 0.0996 ± 0.0155 ⇒ ( 0.0841, 0.1151) .
1014
Based on this interval, we conclude that between 8.41% and 11.51% of the population agree with the
statement “Landlords should have the right to evict a tenant from an apartment because that person
has AIDS.”
9.65 B = 0.1, σ = 0.8
⎡ (1.96 )( 0.8 ) ⎤
2 2
⎡ 1.96σ ⎤
⎥ = (15.68 ) = 245.86
2
n=⎢ ⎥ =⎢
⎣ B ⎦ ⎣ 0.1 ⎦
Since a partial observation cannot be taken, n should be rounded up to n = 246.
9.66 The 99% confidence interval computed from this sample data is
0.71( 0.29 )
0.71 ± 2.58 ⇒ 0.71 ± 0.039 ⇒ ( 0.671, 0.749 ) .
900
Based on this interval, we conclude that between 67.1% and 74.9% of the population of Californians
support allowing 10-2 verdicts in criminal cases not involving the death penalty.
214
9.67 The 99% confidence interval for the mean commuting distance based on this sample is
s ⎛ 6.2 ⎞
x ± (t critical) ⇒ 10.9 ± (2.58) ⎜ ⎟ ⇒ 10.9 ± 0.924 ⇒ ( 9.976, 11.824 ) .
n ⎝ 300 ⎠
9.68 p = 445/602 = 0.739. The 95% confidence interval for the true proportion of California registered
voters who prefer the cigarette tax increase is
p(1 − p) 0.739(0.261)
p± (z critical) ⇒ 0.739 ± 1.96 ⇒ 0.739 ± 0.035 ⇒ ( 0.704, 0.774 ) .
n 602
the distribution of systolic blood pressure of anabolic-steroid-using athletes be approximately like a
normal distribution.
9.69 Example 9.8 has shown that the sample data meets the conditions for the t confidence interval.
99% upper confidence bound for the true mean wait time for bypass patients in Ontario:
s ⎛ 10 ⎞
x + 2.33 = 19 + 2.33 ⎜ ⎟ = 19 + 1.004 = 20.004
n ⎝ 539 ⎠
142
9.70 p= = 0.2801
507
The 95% confidence interval for the proportion of the entire population that could correctly describe
the Bill of Rights as the first ten amendments to the U.S. Constitution is
0.2801(0.7199)
0.2801 ± 1.96 ⇒ 0.2801 ± 1.96(0.0199) ⇒ 0.2801 ± 0.0391 ⇒ (0.2410, 0.3192).
507
9.71 The 95% confidence interval for the population standard deviation of commuting distance is
⎛ 6.2 ⎞
6.2 ± 1.96 ⎜ ⇒ 6.2 ± 0.496 ⇒ (5.704, 6.696).
⎜ 2(300) ⎟⎟
⎝ ⎠
9.72 The width of the interval discussed in the text is

⎛ s ⎞ ⎛ s ⎞ s s
⎜ x + (1.96) ⎟ − ⎜ x − (1.96) ⎟ = 2(1.96) = 3.92 .
⎝ n⎠ ⎝ n⎠ n n
The width of the interval suggested in this problem is

⎛ s ⎞ ⎛ s ⎞ s s
⎜ x + (1.75) ⎟ − ⎜ x − (2.33) ⎟ = (1.75 + 2.33) = 4.08 .
⎝ n⎠ ⎝ n⎠ n n
Since this latter interval is wider (less precise) than the one discussed in the text, its use is not
recommended.
9.73 The t critical value for a 90% confidence interval when df = 12 − 1 = 11 is 1.80. The confidence
interval based on this sample data is
s ⎛ 7.7 ⎞
x ± (t critical) ⇒ 21.9 ± (1.80) ⎜ ⎟ ⇒ 21.9 ± 4.00 ⇒ (17.9, 25.9) .
n ⎝ 12 ⎠
With 90% confidence, the mean time to consume a frog by Indian False Vampire bats is between
17.9 and 25.9 minutes.
215
Chapter 10
10.1 x = 50 is not a legitimate hypothesis, because x is a statistic, not a population characteristic.

Hypotheses are always expressed in terms of a population characteristic, not in terms of a sample
statistic.
10.2 Those that do not comply are parts a, b, d, and e. The reasons for non-compliance are:
a. Both Ho and Ha are identical statements.
b. The hypothesized value on the right side of Ho and Ha must be the same.
d. Same as part b.
e. p is not a population characteristic.
10.3 If we use the hypothesis Ho: μ = 100 versus Ha: μ > 100, we are taking the position that the welds do
not meet specifications, and hence, are not acceptable unless there is substantial evidence to show
that the welds are good (i.e. μ > 100). If we use the hypothesis Ho: μ = 100 versus Ha: μ < 100, we
initially believe the welds to be acceptable, and hence, they will be declared unacceptable only if
there is substantial evidence to show that the welds are faulty. It seems clear that we would choose
the first set-up, which places the burden of proof on the welding contractor to show that the welds
meet specifications.
10.4 a. H0: concealed weapons laws do not reduce crime

Ha: concealed weapons laws do reduce crime
The null hypothesis is the one that has “acceptance until proved otherwise”. The alternative
hypothesis always has the burden of proof.
b. The null hypothesis was not rejected. There was not enough evidence to suggest that
concealed weapons reduce crime. But note! That doesn’t mean there was enough evidence
to prove they increase crime!
10.5 Not being able to conclude that the MMR does cause autism is not the same as concluding that the
MMR does not cause autism – just as failing to reject the null hypothesis is not the same as accepting
the null hypothesis.
10.6 Since the administration has decided to make the change if it can conclude that more than 60% of the
faculty favor the change, the appropriate hypotheses are
Ho: π = 0.6 versus Ha: π > 0.6.
Then, if Ho is rejected, it can be concluded that more than 60% of the faculty favor a change.
10.7 Ho: π = 0.5 versus Ha: π > 0.5
10.8 Ho: μ = 170 Ha: μ < 170
10.9 A majority is defined to be more than 50%. Therefore, the commissioner should test:
Ho: π = 0.5 versus Ha: π > 0.5
10.10 Ho: μ = 7.3 versus Ha: μ > 7.3
10.11 Since the manufacturer is interested in detecting values of μ which are less than 40, as well as values
of μ which exceed 40, the appropriate hypotheses are: Ho: μ = 40 versus Ha: μ =/ 40
217
Exercises 10.12 – 10.22
10.12 a. Thinking that cancer is present when it is not means that the null hypothesis is true but it is
rejected. This is precisely the definition of type-I error.
b. A type-I error occurs in this problem when a cancer screening test result is judged to indicate
presence of cancer even though the patient is actually cancer-free. The main consequence of
such an error is that additional follow-up tests will be performed or treatments for cancer may
be prescribed even though they are not necessary.
c. A type-II error occurs in this problem when a patient actually has cancer but the screening
tests fail to indicate this. As a consequence of such an error the patient will lose the
opportunity to receive timely treatment and may even ultimately die.
d. Given the same amount of data (information), any strategy to lower the risk of a type-I error
will increase the risk of a type-II error.
10.13 a. Type I: Concluding the symptoms are due to a disease when, in truth, the symptoms
are due to child abuse.
Type II: Concluding the symptoms are due to child abuse when, in truth, the symptoms are
due to a disease.
b. Based on the quote, the doctor considers a Type I error to be more serious.
10.14 a. A type I error would be deciding that a particular man is not the father when in fact he is the
father. A type II error would be deciding that a particular man is the father when in fact he is
not the father.
b. α = 0.001 β=0
c. α = 0.001 β = 0.008
10.15 a. Pizza Hut’s decision is consistent with the decision of rejecting Ho.
b. Rejecting Ho when it is true is called a type I error. So if they incorrectly reject Ho, they are
making a type I error.
10.16 a. In this hypothesis test, a type I error consists of saying that the manufacturer’s claim is not
true (π < 0.9), when in fact the manufacturer is correct in its claim. A type II error occurs if the
manufacturer’s claim is incorrect but the consumer agency fails to detect it.
Since a type I error is to accuse the manufacturer of false advertising, the consumer agency
would want to avoid making that error and the possible expensive litigation that would result
from such a false accusation.
A type II error means the manufacturer can continue the false advertising without penalty.
b. Since a type I error has quite serious consequences, the consumer agency may want to use
a small value for α, such as 0.01.
10.17 a. A type I error is returning to the supplier a shipment which is not of inferior quality.
A type II error is accepting a shipment of inferior quality.
b. The calculator manufacturer would most likely consider a type II error more serious, since
they would then end up producing defective calculators.
c. From the supplier’s point of view, a type I error would be more serious, because the supplier
would end up having lost the profits from the sale of the good printed circuits.
218
10.18 A type I error involves concluding that the water being discharged from the power plant has a mean
temperature in excess of 150o F when, in fact, the mean temperature is not greater than 150o F. A
type II error is concluding that the mean temperature of water being discharged is 150o F or less
when, in fact, the mean temperature is in excess of 150o F. I would consider a type II error to be the
more serious. If we make a type II error, then damage to the river ecosystem will occur. Since it
generally takes a long time to repair such damage, I would want to avoid this error. A type I error
means that we require the power plant to take corrective action when it was not necessary to do so.
However, since there are alternative power sources, the consequences of this error are basically
financial in nature.
10.19 a. The manufacturer claims that the percentage of defective flares is 10%. Certainly one would
not object if the proportion of defective flares is less than 10%. Thus, one’s primary concern
would be if the proportion of defective flares exceeds the value stated by the manufacturer.
b. A type I error entails concluding that the proportion of defective flares exceeds 10% when, in
fact, the proportion is 10% or less. The consequence of this decision would be the filing of
charges of false advertising against the manufacturer, who is not guilty of such actions. A
type II error entails concluding that the proportion of defective flares is 10% when, in reality,
the proportion is in excess of 10%. The consequence of this decision is to allow the
manufacturer who is guilty of false advertising to continue bilking the consumer.
10.20 a. Since a high level of mercury content is of particular concern, and the purpose of the test is to
determine if μ is in excess of 5, the appropriate set of hypotheses is
Ho: μ = 5 versus Ha: μ > 5.
b. A type I error is concluding that the mercury content is unacceptably high, when it is not. In
the interest of public safety, choose the larger significance level of 0.01.
10.21 a. They failed to reject the null hypothesis, because their conclusion “There is no evidence of
increased risk of death due to cancer” for those living in areas with nuclear facilities is
precisely what the null hypothesis states.
b. They would be making a type II error since a type I error is failing to reject the null hypothesis
when the null is false.
c. Since the null hypothesis is the initially favored hypothesis and is presumed to be the case
until it is determined to be false, if we fail to reject the null hypothesis, it is not proven to be
true. There is just not sufficient evidence to refute its presumed truth. On the other hand, if
the hypothesis test is conducted with
Ho : π is greater than the value for areas without nuclear facilities
Ha : π is less than or equal to the value for areas without nuclear facilities
and the null hypothesis is rejected, then there would be evidence based on data against the
presumption of an increased cancer risk associated with living near a nuclear power plant.
This is as close as one can come to “proving” the absence of an increased risk using
statistical studies.
10.22 a. The manufacturer should test the hypotheses

Ho: π = 0.02 versus Ha: π < 0.02.
The null hypothesis is implicitly stating that the proportion of defective installations using
robots is 0.02 or larger. That is, at least as large as for humans. In other words, they will not
undertake to use robots unless it can be shown quite strongly that the defect rate is less for
robots than for humans.
b. A type I error is changing to robots when in fact they are not superior to humans. A Type II
error is not changing to robots when in fact they are superior to humans.
219
c. Since a type I error means substantial loss to the company as well as to the human
employees who would become unemployed a small α should be used. Therefore, α = 0.01
is preferred.
Exercises 10.23 – 10.44
10.23 The P-value is the probability, assuming Ho, of obtaining a test statistic value at least as contradictory
to Ho as what actually resulted.
a. A P-value of 0.0003 indicates that these test results are very unlikely to occur if Ho is true.
The more plausible explanation for the occurrence of these results is that Ho is false.
b. A P-value of 0.35 indicates that these results are quite likely to occur (consistent with) if Ho is
true. The data do not cast reasonable doubt on the validity of Ho.
10.24 Ho is rejected if P-value ≤ α . Since α = 0.05, Ho should be rejected for the following P-values:
a: 0.001, b: 0.021, and d: 0.047
10.25 Ho is rejected if P-value ≤ α . Ho should be rejected for only the following pair:
d: P-value = 0.084, α = 0.10
10.26 Using Table 2 from the Appendix we get the following:
a. 0.0808
b. 0.1762
c. 0.0250
d. 0.0071
e. 0.5675
10.27 a. nπ = 25(0.2) = 5. nπ ≤ 10, The large sample test is not appropriate.
b. nπ = 10(0.6) = 6. nπ ≤ 10, The large sample test is not appropriate.
c. nπ = 100(0.9) = 90, n(1 − π) = 100(0.1) = 10, Both ≥ 10. The large sample test is
` appropriate.
d. nπ = 75(0.05) = 3.25. nπ ≤ 10, The large sample test is not appropriate.
10.28 Let π represent the proportion of American adults who oppose reinstatement of the draft
Ho: π = 0.667
Ha: π > 0.667
We will compute a P-value for this test. α = 0.05.
Since the population is much larger than the sample, nπ = 1000(0.667) ≥ 10, and n(1 − π) =
1004(0.333) ≥ 10, the given sample is a random sample and the sample size is large, the large
sample z test may be used.
n = 1000, p = 700/1000 = 0.7
220
0.7 − 0.667
z= = 2.21
0.667(0.333)
1000
P-value = area under the z curve to the right of 2.21 = 0.0136
Since the P-value is less than α, H 0 is rejected.
There is enough evidence to suggest that more than two-thirds of all American adults oppose
reinstatement of the draft.
10.29 Let π represent the proportion of adults Americans who would favor the drafting of women.
Ho: π = 0.5
Ha: π < 0.5
Since the population is much larger than the sample, nπ = 1000(0.5) ≥ 10, and n(1 − π) = 1000(0.5) ≥
10, the given sample is a random sample and the sample size is large, the large sample z test may
be used.
n = 1000, p = .43
0.43 − 0.5
z= = − 4.43
0.5(0.5)
1000
P-value = area under the z curve to the left of -4.43 ≈ 0
There is enough evidence to suggest that less than half of all adult Americans would favor the
drafting of women.
10.30 Let π represent the proportion cell phone users who have received commercial messages or ads in
2004.
Ho: π = 0.13
Ha: π > 0.13
We will compute a P-value for this test. Even though the problem doesn’t state what value is to be
used for α, for illustrative purposes we will use α = 0.05.
5500(0.87) ≥ 10, the given sample is a random sample and the sample size is large, the large sample
z test may be used.
n = 5500, p = .2
0.2 − .13
z= = 15.44
.13(.87)
5500
221
P-value = area under the z curve to the right of 15.44 ≈ 0
There is enough evidence to suggest that the proportion cell phone users who have received
commercial messages or ads in 2004 is greater than the proportion of .13 reported for the previous
year.
10.31 Let π represent the proportion of adult Americans who plan to alter their shopping habit if gas prices
remain high.
Ho: π = 0.75
Ha: π > 0.75
Even though the problem doesn’t state what value is to be used for α, for illustrative purposes we will
use α = 0.05
1813(0.25) ≥ 10, the given sample is a random sample and the sample size is large, the large sample
z test may be used.
n = 1813, p = 1432/1813 = 0.79
0.79 − .75
z= = 3.93
0.75(0.25)
1813
Since the P-value is less than α of 0.01, H 0 is rejected.
There is enough evidence to suggest that more than three-quarters of adult Americans plan to alter
their shopping habits if gas prices remain high.
10.32 Let π represent the proportion of U.S. adults who say they would not be bothered if the NSA collected
records of personal telephone calls they had made.
Ho: π = 0.50
Ha: π > 0.50
p − hypothesized value p − .5
Test Statistic: z = =
(hypothesized value)(1 - hypothesized value) (.5)(.5)
n n
Assumptions: This test requires a random sample and a large sample size. The given sample was a
random sample, the population size is much larger than the sample size, and the sample size was n =
502. Because 502(.5) ≥ 10 and 502(1-.5) ≥ 10, the large-sample test is appropriate.
n = 502, p = 331/502 = .66
222
0.66 − 0.5
z= = 7.17
0.5(0.5)
502
The data provides enough evidence to suggest that more than half of all U.S. adults say they would
not be bothered if the NSA collected records of personal telephone calls they had made.
10.33 Let π represent the proportion of U.S. adults who believe that playing the lottery is the best strategy
for accumulating $200,000 in net wealth.
Ho: π = 0.2
Ha: π > 0.2
use α = 0.05
n n
1000. Because 1000 (.2) ≥ 10 and 1000(.8) ≥ 10, the large-sample test is appropriate.
n = 1000, p = .21
0.21 − 0.2
z= = 0.79
0.2(0.8)
1000
Since the P-value is greater than α, H 0 is not rejected.
The data does not provide enough evidence to suggest that more than 20% of all adult Americans
believe that playing the lottery is the best strategy for accumulating $200,000 in net wealth.
10.34 Let π represent the proportion of adult Americans who prefer to watch movies at home.
Ho: π = 0.5
Ha: π > 0.5
α = 0.05
n n
223
1000. Because 1000 (.5) ≥ 10 and 1000(1-.5) ≥ 10, the large-sample test is appropriate.
n = 1000, p = .73
0.73 − 0.5
z= = 14.55
0.5(0.5)
1000
The data provides enough evidence to suggest that the majority of American adults would prefer to
watch movies at home rather than at a movie theater.
10.35 a. Let π represent the proportion of adult Americans who think that the quality of movies being
produced is getting worse.
Ho: π = 0.5
Ha: π < 0.5
α = 0.05
n n
Assumptions: This test requires a random sample and a large sample size. The given sample
was a random sample, the population size is much larger than the sample size, and the
sample size was n = 1000. Because 1000 (.5) ≥ 10 and 1000(1-.5) ≥ 10, the large-sample
test is appropriate.
n = 1000, p = .47
0.47 − 0.5
z= = −1.90
0.5(0.5)
1000
P-value = area under the z curve to the left of -1.90 = .0287
Since the P-value is less than α, Ho is rejected.

The data provides enough evidence to suggest that fewer than half of American adults
believe that movie quality is getting worse.
b. The hypothesis test would be the identical up to the assumptions. The sample size is 100.
Because 100(.5) ≥ 10 and 100(1-.5) ≥ 10, the large-sample test is appropriate.
n = 100, p = .47
0.47 − 0.5
z= = −0.6
0.5(0.5)
100
224
P-value = area under the z curve to the left of -0.6 = .2743

The data does not provide enough evidence to suggest that fewer than half of American
adults believe that movie quality is getting worse.
(hypothesized value)(1 - hypothesized value)

c. The denominator, σ p : has the same
n
hypothesized value for both but for part (a), n = 1000, and for part (b), n = 100. The value of
the test statistic for part (b) is larger which leads to a larger P-value. The P-value in (a) is
smaller than α and the P-value in (b) is larger than α which leads to different conclusions.
With a small sample the difference between the sample proportion of 0.47 and the
hypothesized proportion of 0.5 could plausibly be attributed to chance, whereas when the
sample size is larger this difference is no longer likely to be attributable to chance.
10.36 a. Let π represent the proportion of U.S. businesses monitor employees’ web site visits.
Ho: π = 0.75
Ha: π > 0.75
α = 0.01

n n
sample size was n = 526. Because 526(.75) ≥ 10 and 526(.25) ≥ 10, the large-sample test is
appropriate.
n = 526, p = 400/526 = .76
0.76 − 0.75
z= = 0.53
0.75(0.25)
526
P-value = area under the z curve to the right of 0.53 = .2981

The data does not provide enough evidence to suggest that more than 75% of U.S.
businesses monitor employees’ web sites.
b. Let π represent the proportion of U.S. businesses monitor employees’ web site visits.
Ho: π = 0.5
Ha: π > 0.5
α = 0.01
n n
225
sample size was n = 526. Because 526(.5) ≥ 10 and 526(1 - .5) ≥ 10, the large-sample test is
appropriate.
n = 526, p = 400/526 = .76
0.76 − 0.5
z= = 11.93
0.5(0.5)
526

The data provides enough evidence to suggest that the majority of U.S. businesses monitor
employees’ web sites.
10.37 Let π represent the proportion of all baseball fans that think that the designated hitter rule should
either be expanded to both baseball leagues or eliminated.
Ho: π = 0.5
Ha: π > 0.5
use α = 0.05
n n
random sample, the population size is much larger than the sample size, and the sample size was
n =394. Because 394(.5) ≥ 10 and 394(1 - .5) ≥ 10, the large-sample test is appropriate.
n = 394, p = 272/394 = .69
0.69 − 0.5
z= = 7.54
0.5(0.5)
394

The data provides enough evidence to suggest that the majority of all baseball fans think that the
designated hitter rule should either be expanded to both baseball leagues or eliminated.
10.38 Let π represent the proportion of all adult Americans who can name at least one justice who is
currently serving on the U.S. Supreme Court.
Ho: π = 0.5
Ha: π < 0.5
α = 0.01
226
n n
n = 1000, p = .43
0.43 − 0.5
z= = −4.43
0.5(0.5)
1000

The data provides enough evidence to suggest that fewer than half of all adult Americans can name
at least one justice currently serving on the Supreme Court.
10.39 Let π represent the proportion of U.S. adults who believe that rudeness is a worsening problem.
Ho: π = 0.75
Ha: π > 0.75
Since nπ = 2013(0.75) ≥ 10, and n(1 − π) = 2013(0.25) ≥ 10, the large sample z test may be used.
n = 2013, p = 1283/2013 = 0.637357
0.63736 − 0.75 −.11264

z= = = − 11.67
0.75(0.25) .00965
2013
P-value = area under the z curve to the right of –11.67 ≈ 1
Since the P-value is greater than α, Ho cannot be rejected.
There is not enough evidence to suggest that over ¾ of U.S. adults think that rudeness is a
worsening problem.
10.40 Let π represent the proportion of U.S. adults who believe that the Census Bureau will keep
information confidential.
Ho: π = 0.5
Ha: π < 0.5
n = 1004, p = 432/1004 = 0.43028
227
0.43208 − 0.5 −.06792
z= = = − 4.3042
0.5(0.5) .01578
1004
P-value = area under the z curve to the left of –4.3 ≈ 0
There is enough evidence to suggest that fewer than half of U.S. adults believe that the Census
Bureau will keep information confidential.
10.41 Let π represent the proportion of religion surfers who belong to a religious community
Ho: π = 0.68
Ha: π ≠ 0.68
n = 512, p = .84
0.84 − .68 .16

z= = = 7.77
0.68(0.32) .0206
512
P-value = 2(area under the z curve to the right of 7.77) ≈ 2(0) = 0
There is enough evidence to suggest that the proportion of religion surfers that belong to a religious
community is different than 68%.
10.42 Let π represent the proportion of accidents involving teenage drivers.
Ho: π = 0.07
Ha: π ≠ 0.07
We will compute a P-value for this test. Even though the problem doesn’t state what value is to be
used for α, for illustrative purposes we will use α = 0.05.
n = 500, p = .14
0.14 − .07 .07
z= = = 6.14
0.07(0.93) .0114
500
P-value = 2(area under the z curve to the right of 6.14) ≈ 2(0) = 0
There is enough evidence to suggest that the proportion of accidents involving teenage drivers is
different from 0.07.
228
10.43 Let π represent the proportion of the time the Belgium Euro lands with its head side up.
Ho: π = 0.5
Ha: π ≠ 0.5
n = 250, p = 140/250 = 0.56
0.56 − .5 .06
z= = = 1.90
0.5(0.5) .0316
250
P-value = 2(area under the z curve to the right of 1.90) ≈ 2(0.0287) = 0.0574
Since the P-value is greater than α of 0.01, Ho is not rejected.
There is not enough evidence to suggest that the proportion of the time that the Belgium Euro coin
would land with its head side up is not 0.5. With a significance level of 0.05, the same
conclusion would be reached, since the p-value would still be greater than α.
10.44 The value of “just over 38%” is a population proportion. Inference from a sample is not necessary if
the value of the population is already known.
Exercises 10.45 – 10.64
10.45 Since this is a two-tailed test, the P-value is equal to twice the area captured in the tail in which z
falls. Using Appendix Table 2, the P-values are:
a. 2(0.0179) = 0.0358
b. 2(0.0401) = 0.0802
c. 2(0.2810) = 0.5620
d. 2(0.0749) = 0.1498
e. 2(0) = 0
10.46 Population characteristic of interest: μ = average percentage of silicon dioxide.
Ho: μ = 5
Ha: μ ≠ 5
α = 0.01
x −5
Test statistic: z =
s
n
Since the sample is reasonable large (≥ 30) independent and randomly selected, it is reasonable to
use the z test.
229
Computations: n = 36, x = 5.21, s = 0.38,
5.21 − 5
z= = 3.32
0.38
36
P-value = 2(area under the z curve to the right of 3.32) = 2(1 − 0.9995) = 0.0010
Since the P-value of 0.0010 is less than the α of 0.01, Ho is rejected at the 0.01 level. There is
sufficient evidence to conclude that the true mean percentage of silicon dioxide in this type of cement
has a value which differs from 5.
10.47 a. P-value = area under the 8 d.f. t curve to the right of 2.0 = 0.040
b. P-value = area under the 13 d.f. t curve to the right of 3.2 = 0.003
c. P-value = area under the 10 d.f. t curve to the left of −2.4

= area under the 10 d.f. t curve to the right of 2.4 = 0.019
d. P-value = area under the 21 d.f. t curve to the left of −4.2

e. P-value = 2(area under the 15 d.f. t curve to the right of 1.6) = 2 (0.065) = 0.13
f. P-value = 2(area under the 15 d.f. t curve to the right of 1.6) = 2 (0.065) = 0.13
g. P-value = 2(area under the 15 d.f. t curve to the right of 6.3) = 2(0) = 0
10.48 a P-value = 2(area under the 9 d.f. t curve to the right of 0.73)
2(0.251) > P-value > 2(0.222) ⇒ 0.502 > P-value > 0.444
b. P-value = area under the 10 d.f. t curve to the right of −0.5

= 1 − area under the 10 d.f. t curve to the right of 0.5 = 1 − 0.314 = 0.686
c. P-value = area under the 19 d.f. t curve to the left of −2.1

d. P-value = area under the 19 d.f t curve to the left of −5.1

= area under the 19 d.f. t curve to the right of 5.1 = 0
e. P-value = 2(area under the 39 d.f. t curve to the right of 1.7) ≈ 2(0.048) = 0.096
10.49 The P-value for this test is equal to the area under the 14 d.f. t curve to the right of 3.2 = 0.003.
a. α = 0.05, reject Ho
b. α = 0.01, reject Ho
c. α = 0.001, fail to reject Ho
10.50 a. P-value = area under the 17 d.f. t curve to the left of −2.3 = 0.017.
Since 0.017 is less than 0.05, Ho is rejected.
b. P-value = area under the 17 d.f. t curve to the left of −1.83.

Thus, 0.045 > P-value > 0.037. Since the P-value is greater than 0.037, which is greater than
α = 0.01, Ho is not rejected.
230
c. P-value = area under the 17 d.f. t curve to the left of 0.47
= 1 − area under the 17 d.f. t curve to the right of 0.47 ≈ 1 − 0.322 = 0.678.
Since the P-value exceeds any reasonable choice of α, Ho is not rejected.
10.51 a. P-value = 2(area under the 12 d.f. t curve to the right of 1.6) = 2(0.068) = 0.136.
Since P-value > α, Ho is not rejected.
b. P-value = 2(area under the 12 d.f. t curve to the left of −1.6)

= 2(area under the 12 d.f. t curve to the right of 1.6) = 2(0.068) = 0.136.
c. P-value = 2(area under the 24 d.f. t curve to the left of −2.6)

d P-value = 2(area under the 24 d.f. t curve to the left of −3.6)

Ho would be rejected for any α > 0.002.
10.52 Let μ = mean number of credit cards that undergraduates report carrying.
H 0 : μ = 4.09
H a : μ < 4.09
α = 0.01 (A value for α was not specified in the problem. The value 0.01 was chosen for illustration.)
x − hypothesized value x − 4.09

Test statistic: t = =
s s
n n
This test requires a random sample and either a large sample or a normal population distribution.
Because the sample is large (n = 132), it is reasonable to proceed with the t test as long as we are
willing to consider the undergraduates in the study as if they were a random sample from the
population of all undergraduates.
2.6 − 4.09
n = 132, x = 2.6, s = 1.2. t= = −14.3
1.2
132
P-value = area under the 131 d.f. t curve to the left of -14.3 ≈ 0
Since the P-value is less than α, the null hypothesis is rejected at the 0.01 level. The data supports
the conclusion that the mean number of credit cards that all undergraduates would report carrying is
less than the credit bureau’s figure of 4.09 cards.
10.53 Let μ = mean wrist extension while using the new mouse design.
H 0 : μ = 20
H a : μ > 20
α = 0.05 (A value for α was not specified in the problem. The value 0.05 was chosen for illustration.)
231
x − hypothesized value x − 20
s s
n n
This test requires a random sample and either a large sample or a normal population distribution. If
this sample was a random sample of Cornell University students, it would be appropriate to
generalize the results of the test to the population of Cornell students. However, if we wanted to
generalize the results of this study to all university students, we would have to assume that the
sample of the 24 Cornell students were a random sample of all university students, which they are
clearly not. From the boxplot (below), the data is symmetrical with no outliers, it is not unreasonable
to assume that the sample came from a normal distribution.
22 23 24 25 26 27 28 29 30 31
extension (degrees)
25.9 − 20
n = 24, x = 25.9, s = 1.95. t= = 14.8
1.95
24
P-value = area under the 23 d.f. t curve to the right of 14.8 ≈ 0
the conclusion that the mean wrist extension for all Cornell undergraduates using this mouse design
is greater than 20 degrees.
10.54 Let μ = mean minimum purchase amount (in $) for which Canadians consider the use of a debit card
to be appropriate.
H 0 : μ = 10
H a : μ < 10
α = 0.01
s s
n n
The Canadians were randomly selected and because the sample is large (n = 2000), it is reasonable
to proceed with the t test.
232
9.15 − 10
n = 2000, x = 9.15, s = 7.6. t = = −5.0
7.6
2000
the conclusion that the mean minimum purchase amount for which Canadians consider the use of a
debit card is less than $10.
10.55 A hypothesis test will give one of two results: either statistically significant or not statistically
significant; it won’t report any quantity or difference. A confidence interval is needed for that
information. It is possible for a result to be significantly significant but at such a small difference from
the null hypothesis that from a practical point of view, the difference has no practical consequence.
10.56 a. Let μ = mean time (in hours per week) spent using the Internet by Canadians
H 0 : μ = 12.5
H a : μ > 12.5
α = 0.05

s s
n n
This test requires a random sample and either a large sample or a normal population
distribution. The Canadians were randomly selected and because the sample is large (n =
1000), it is reasonable to proceed with the t test.
12.7 − 12.5
n = 1000, x = 12.7, s = 5. t= = 1.26
5
1000
P-value = area under the 999 d.f. t curve to the right of 1.26 ≈ .1038
Since the P-value is greater than α, the null hypothesis cannot be rejected at the 0.01 level.
The data does not support the conclusion that the mean time spent using the Internet by
Canadians is greater than 12.5 hours per week.
b. Let μ = mean time (in hours per week) spent using the Internet by Canadians
H 0 : μ = 12.5
H a : μ > 12.5
α = 0.05

s s
n n
233
distribution. The Canadians were randomly selected and because the sample is large (n =
12.7 − 12.5
n = 1000, x = 12.7, s = 2. t = = 3.16
2
1000
P-value = area under the 999 d.f. t curve to the right of 3.16 ≈ .0008
Since the P-value is less than α, the null hypothesis is rejected at the 0.01 level. The data
supports the conclusion that the mean time spent using the Internet by Canadians is greater
than 12.5 hours per week.
c. Because the sample standard deviation was smaller in Part (b), and everything else stayed
the same, the t statistic was larger which led to a smaller P-value. The smaller variability in
the sample made the difference more obvious.
10.57 a. The large standard deviation means that the distribution for online times for the teens had a
lot of variability. As it is large compared to the mean, it is likely that the distribution is
heavily skewed to the left.
b. Let μ = mean number of hours per week that teens spend online.
H 0 : μ = 10
H a : μ > 10
α = 0.05. (A value for α was not specified in the problem. The value 0.05 was chosen for
illustration.)
s s
n n
distribution. The teens were randomly selected and although the distribution of the data is
probably skewed, because the sample is large (n = 534), it is reasonable to proceed with the t
test.
14.6 − 10
n = 534, x = 14.6, s = 11.6. t= = 9.16
11.6
534
P-value = area under the 533 d.f. t curve to the right of 9.16 ≈ 0
supports the conclusion that the mean time spent online by teens is greater than 10 hours per
week.
234
10.58 a. Let μ = mean number of hours per week that parents thought their Canadian teens spend
online.
H 0 : μ = 10
H a : μ < 10
α = 0.05. (A value for α was not specified in the problem. The value 0.05 was chosen for
illustration.)
s s
n n
distribution. The Canadian teens were randomly selected because the sample is large (n =
6.5 − 10
n = 676, x = 6.5, s = 8.6. t= = −10.58
8.6
676
supports the conclusion that the mean time that parents think their Canadian teens spend
online is less than 10 hours per week.
b. Both questions looked at the mean number of hours per week that the teens spent online.
However for 10.57 b, the data was reported by the teens themselves, and for 10.58a, the
data was reported by their parents. Obviously the parents grossly underestimate the amount
of time the teens spend online!
10.59 The authors are saying that the positive effects of music on pain intensity found in the study are
statistically significant, but not practically significant from a clinical point of view.
10.60 Let μ denote the true average “speaking up” value for Asian men.
Ho: μ = 10
Ha: μ < 10
α = 0.05 (A value for α was not specified in the problem, so this value was chosen.)
use the t test.
x − 10
Test statistic: t =
s
n
235
8.75 − 10
t= = − 3.89
2.57
64
P-value = area under the 63 d.f. t curve to the left of -3.89 ≈ 0.
Conclusion: Since the P-value is less than α, Ho is rejected. The data in this sample does support
the conclusion that the average “speaking up” score for Asian men is smaller than 10.0.
10.61 Population characteristic of interest: μ = population mean MWAL
Ho: μ = 25
Ha: μ > 25
α = 0.05
x − 25
Test statistic: t =
s
n
Assumptions: This test requires a random sample and either a large sample size (generally n ≥ 30) or
a normal population distribution. Since the sample size is only 5, and a box plot of the data does not
show perfect symmetry, does this seem reasonable? Based on the their understanding of MWAL
values, the authors of the article thought it was reasonable to assume that the population distribution
was approximately normal and based on their expert judgment, we will proceed, with caution, with the
t-test.
Computations: n = 5, x = 27.54 , s = 5.47
27.54 − 25
t= = 1.0
5.47
5
P-value = area under the 4 d.f. t curve to the right of 1 = 0.187
Conclusion: Since the P-value is greater than α, the null hypothesis is not rejected at the 0.01 level.
There is not enough evidence to suggest that the mean MWAL exceeds 25.
10.62 Let μ denote the true average attention span (in minutes) of teenage Australian boys.
Ho: μ = 5
Ha: μ < 5
α = 0.01
use the t test.
x −5
Test statistic: t = with d.f. = 50 − 1 = 49
s
n
Computations: n = 50, x = 4, s = 1.4
236
4−5 −1
t= = = −5.0508
1.4 0.198
50
P-value = area under the 49 d.f. t curve to the left of –5.0508 ≈ 0.0
Since the P-value is smaller than α, Ho is rejected. The sample does provide convincing evidence
that the average attention span of Australian teenagers is less than 5 minutes.
10.63 a. Since the boxplot is nearly symmetric and the normal probability plot is very much like a
straight line, it is reasonable to use a t-test to carry out the hypothesis test on μ.
b. The median is slightly less than 245 and because of the near symmetry, the mean should be
close to 245. Also, because of the large amount of variability in the data, it is quite
conceivable that the average calorie content is 240.
c. Let μ denote the true average calorie content of this type of frozen dinner.
Ho: μ = 240
Ha: μ =/ 240
α = 0.05 (A value for α was not specified in the problem, so this value was chosen).
Since the sample is reasonable large (≥ 30) independent and randomly selected, it is
reasonable to use the t test.
x − 240
s
n
Computations: n = 12, x = 244.333, s = 12.383
244.333 − 240 4.333

t= = = 1.21
12.383 3.575
12
P-value = 2(area under the 11 d.f. t curve to the right of 1.21) ≈ 2 (0.128) = 0.256
Since the P-value exceeds α, Ho is not rejected. The sample evidence does not support the
conclusion that the average calorie content differs from 240.
10.64 Let μ denote the true average rate of uptake of radio-labeled amino acid when grown in a medium
containing nitrates.
Ho: μ = 8000
Ha: μ < 8000
α = 0.10
Assuming that rate of uptake has a normal distribution, the test statistic is
x − 8000
t= with d.f. = 15 − 1 = 14
s
n
From the data given, n = 15, x = 7788.8, s = 1002.4308.
237
7788.8 − 8000
t= = − 0.816
1002.4308
15
P-value = area under the 14 d.f. t curve to the left of −0.816 = 0.214.
Since the P-value exceeds α, the null hypothesis is not rejected. It cannot be concluded that the
addition of nitrates results in a decrease in true average uptake.
Exercises 10.65 – 10.71
10.65 a. When the significance level is held fixed, increasing the sample size will increase the power
of a test.
b. When the sample size is held fixed, increasing the significance level will increase the power
of a test.
10.66 a. The z statistic given in this problem is an appropriate test statistic in this setting because:
i. The parameter being tested is a population mean.
ii. The variance of the population being sampled is assumed known.
iii. The sample size is sufficiently large, and hence by the central limit theorem, the
distribution of the random variable x should be approximately normal.
b. A type I error involves concluding that the water being discharged from the power plant has a
mean temperature in excess of 150o F when, in fact, the mean temperature is not greater
than 150o F. A type II error is concluding that the mean temperature of water being
discharged is 150o F or less when, in fact, the mean temperature is in excess of 150o F.
c. From Appendix Table 2, the area to the right of 1.8 is 0.0359. Hence, rejecting Ho when
z > 1.8 corresponds to an α value of 0.0359.
x − 150 ⎛ 10 ⎞
d. If z > 1.8, then > 1.8, and it follows that x > 150 + 1.8 ⎜ ⎟ = 152.546.
10 ⎝ 50 ⎠
50
In the graph below, the shaded area = P(type II error when μ = 153)
0.3
Sample mean = 152.546

Area of shaded (corresponds to z = 1.8)
0.2 region = probability
Density
of type II error
0.1
0.0
148 153 158

Sample mean value
238
⎡ ⎤
⎢ 152.546 − 153 ⎥
e. β (when μ = 153) = P( x < 152.546) = P ⎢ z < ⎥ = P(z < −0.32) = 0.3745
⎢ 10 ⎥
⎢⎣ 50 ⎥⎦
⎡ ⎤
⎢ 152.546 − 160 ⎥
f. β (when μ = 160) = P( x < 152.546) = P ⎢ z < ⎥ = P(z < −5.27) ≈ 0
⎢ 10 ⎥
⎢⎣ 50 ⎥⎦
g. From part d, Ho is rejected if x > 152.546. Since x = 152.4, Ho is not rejected. Because Ho
is not rejected, a type II error might have been made.
10.67 a. α = area under the z curve to the left of −1.28 = 0.1003.
b. The decision rule is reject H0 if x < 10 − 1.28(0.1) ⇒ x < 9.872.
9.872 − 9.8
z= = 0.72
0.1
(β when μ = 9.8) = area under the z curve to the right of 0.72

= 1 − area under the z curve to the left of 0.72
= 1 − 0.7642 = 0.2358.
This means that if μ = 9.8, about 24% of all samples would result in x values greater than
9.872 and the nonrejection of Ho : μ = 10.
c. β when μ = 9.5 would be smaller than β when μ = 9.8.
9.872 − 9.5
z= = 3.72
0.1
(β when μ = 9.5) = 1− area under the z curve to the left of 3.72 ≈ 1 − 1 = 0
d. When μ = 9.8, the value of β is 0.2358 from part b. So, power of the test = 1-β
= 1-0.2358 = 0.7642. When μ = 9.5, the value of β is (approximately) 0 from part c. So,
power of the test = 1-β = 1.
10.68 a. Let π be the true proportion of apartments which prohibit children.
Ho: π = 0.75
Ha: π > 0.75
α = 0.05
Since nπ = 125(0.75) = 93.75 ≥ 10, and n(1 − π) = 125(0.25) = 31.25 ≥ 10, the large sample z
test for π may be used.
p− 0.75 0.816 − 0.75

z= = = 1.71
0.75(0.25) 0.0387
125
239
P-value = area under the z curve to the right of 1.71 = 1 − 0.9564 = 0.0436.
Since the P-value is less than α, Ho is rejected. This 0.05 level test does lead to the
conclusion that more than 75% of the apartments exclude children.
p − 0.75
b. The test with α = 0.05 rejects Ho if > 1.645, which is equivalent to
0.0387
p > 0.75 + 0.0387(1.645) = 0.8137. Ho will then not be rejected if p ≤ 0.8137.
When π = 0.80 and n = 125, β = P(not rejecting Ho when π = 0.8) = P(p ≤ 0.8137)
0.8137 − 0.8
= area under the z curve to the left of
0.8(0.2)
125
= area under the z curve to the left of 0.38 = 0.6480.
10.69 a. Let μ denote the mean amount of shaft wear after a fixed mileage.
Ho: μ = 0.035
Ha: μ > 0.035
α = 0.05
x − 0.035
t= with d.f. = 6
s
n
0.0372 − 0.035
t= = 0.466
0.0125
7
P-value = area under the 6 d.f. t curve to the right of 0.466 ≈ 0.329.
Since the P-value exceeds α, Ho is not rejected. It cannot be concluded that the mean
amount of shaft wear exceeds 0.035 inches.
alternative value - hypotheszed value 0.04 − 0.035

b. d= = = 0.4
σ 0.0125
From Appendix Table 5, use the set of curves for α = 0.05, one-tailed test. Enter the table
using d = 0.4, and go up to where the d = 0.4 line intersects with the curve for d.f. = 6. Then
read the β value of the vertical axis. This leads to a β of about 0.75.
c. From b, value of β is about 0.75. So, power of the test = 1-β ≈ 1-0.75 = 0.25.
alternativevalue − hypothesized value

10.70 a. d=
σ
52 − 50
i. d= = 0.2 From Appendix Table 5, β ≈0.85.
10
240
55 − 50
ii. d= = 0.5 From Appendix Table 5, β ≈0.55.
10
60 − 50
iii. d= =1 From Appendix Table 5, β ≈ 0.10.
10
70 − 50
iv. d= = 2 From Appendix Table 5, β ≈ 0.
10
b. As σ increases, d decreases in value. Therefore, the value of β increases.
0.52 − 0.5
10.71 a. d= = 1 From Appendix Table 5, β ≈0.06.
0.02
0.48 − 0.5
b. d= = 1 From Appendix Table 5, β ≈0.06.
0.02
0.52 − 0.5
c. d= = 1 From Appendix Table 5, β ≈0.21
0.02
0.54 − 0.5
d. d= =2 From Appendix Table 5, β ≈ 0.
0.02
0.54 − 0.5
e. d= = 1 From Appendix Table 5, β ≈0.06.
0.04
0.54 − 0.5
f. d= = 1 From Appendix Table 5, β ≈0.01.
0.04
g. Answers will vary from student to student.
Exercises 10.72 – 10.95
10.72 Let π represent the proportion of all smokers who believe that they are at increased risk of cancer.
Ho: π = 0.5
Ha: π < 0.5
α = 0.05
n n
n = 737, p = 295/737 = .40.
0.4 − 0.5
z= = −5.43
0.5(0.5)
737
241
There is virtually no chance of seeing a sample proportion and corresponding z value as extreme as
a result of chance variation alone when H 0 is true. The evidence to suggest that the true proportion
of smokers who view themselves as being at increased risk of cancer is less than 50% is very
compelling.
Note that the random sample was taken from U.S. households with telephones, so any results can
only be generalized to the population of U.S. households with telephones.
10.73 Let π represent the true proportion of all U.S. adults who approve of casino gambling.
Ho: π = 2/3 = .667

Ha: π > .667
α = 0.05 (A value for α was not specified in the problem. The value 0.05 was chosen for
illustration.)
n n
n =1523. Because 1523(.667) ≥ 10 and 1523(.333) ≥ 10, the large-sample test is appropriate.
n = 1523, p = 1035/1523 = .68.
0.680 − 0.667
z= = 1.08
0.667(0.333)
1523
P-value = area under the z curve to the right of 1.08 = .1401

Since the P-value is greater than α, H 0 cannot be rejected.
It would not be all that unusual to observe a sample proportion as large as .680 if the null hypothesis
was true. There is not strong evidence that the proportion of all U.S. adults who favor casino
gambling is great than two-thirds.
10.74 Let π represent the proportion of college students that carry a credit card balance from month to
month.
Ho: π = 0.5
Ha: π > 0.5
α = 0.05.
(hypothesized value)(1 - hypothesized value) (.5)(1 − .5)
n n
242
n = 310, p = 217/310 = 0.7.
0.7 − 0.5
z= = 7.04
0.5(1 − .5)
310
P-value = area under the z curve to the right of 7.04 ≈ 0.00
There is strong evidence that the proportion of all college students who carry a credit card balance
from month to month is greater than .5. The sample provides convincing evidence that the industry
claim is not correct.
10.75 Let π represent the proportion of APL patients receiving arsenic who go into remission
Ho: π = 0.15
Ha: π > 0.15
n = 100, p = 0.42
0.42 − .15 .27

z= = = 7.56
0.15(0.85) .0357
100
Since the P-value is less than α , Ho is rejected.
There is enough evidence to suggest that the proportion of APL patients receiving arsenic who go
into remission is greater than 0.15
10.76 a. Let π represent the proportion of 18-19 year olds who have been asked to buy cigarettes for
an underage smoker.
Ho: π = 0.50
Ha: π < 0.50
We will compute a P-value for this test. Even though the problem doesn’t state what value is
to be used for α, for illustrative purposes we will use α = 0.05.
n = 149, p = 0.436
0.436 − 0.5 −0.064

z= = = − 1.5624
0.5(0.5) 0.04096
149
P-value = area under the z curve to the left of –1.5624 ≈ 0.0591

Since the P-value is greater than α, Ho cannot be rejected. However, if α = 0.10 is used, then
Ho will be rejected.
243
b. Let π represent the proportion of nonsmoking 18-19 year olds who have been asked to buy
cigarettes for an underage smoker.
Ho: π = 0.50
Ha: π < 0.50
We will compute a P-value for this test. Even though the problem doesn’t state what value is
to be used for α, for illustrative purposes we will use α = 0.05.
n = 110, p = 0.382
0.382 − 0.5 −0.118

z= = = − 2.475
0.5(0.5) 0.04767
110
Since the P-value is less than α, Ho can be rejected.
There is sufficient evidence (even with α = 0.01) to support the claim that less than half of
nonsmoking 18-19 year olds have been approached to buy cigarettes.
10.77 Let π represent the proportion of U.S. adults who are aware that an investment of $25 a week could
result in a sum of over $100,000 over 40 years.
Ho: π = 0.4
Ha: π < 0.4
n = 1010, p = 374/1010 = 0.3703
0.3703 − .4 −0.0297
z= = = − 1.93
0.4(0.6) 0.0154
1010
Since the P-value is smaller than α , Ho is rejected.
There is not enough evidence to suggest that proportion of U.S. adults who are aware that an
investment of $25 a week could result in a sum of over $100,000 over 40 years is less than 40%
10.78 Let π represent the proportion of U.S. adults who see lottery or sweepstakes win as their best chance
at accumulating $500,000.
Ho: π = 0.25
Ha: π > 0.25
α = 0.01
244
n = 1010, p = 0.28
0.28 − 0.25 0.03

z= = = 2.2018
0.25(0.75) 0.01363
1010
P-value = area under the z curve to the right of 2.2018 = 0.0138.
Since the P-value is greater than α, Ho cannot be rejected.
There is insufficient evidence (at α = 0.01) to conclude that more than 25% of the U.S. adults see a
lottery or sweepstakes win as their best chance at accumulating $500,000.
10.79 Let π represent the proportion of social sciences and humanities majors who have a B average going
into college but end up with a GPA below 3.0 at the end of their first year.
Ho: π = 0.50
Ha: π > 0.50
We will compute a P-value for this test. The problem does not specify a value for α but we will use
α = 0.05 for illustration.
n = 137, p = 0.532
0.532 − 0.5 0.032

z= = = 0.7491
0.5(0.5) 0.0427
137
P-value = area under the z curve to the right of 0.7491= 0.2269.
Since the P-value is greater than α, Ho cannot be rejected at the level of significance of 0.05.
The data does not support the conclusion that a majority of students majoring in social sciences and
humanities who enroll with a HOPE scholarship will lose their scholarship.
10.80 Population characteristic of interest: μ = true mean weight of starters on teams that are not ranked in
the top 20.
Ho: μ = 105
Ha: μ < 105
α = 0.05
x − 105
Test statistic: t =
s
n
a normal population distribution.
245
103.3 − 105 −1.7
t= = = − 0.60
16.3 2.838
33
P-value = area under the 32 d.f. t curve to the left of -0.60 ≈ 0.277
Conclusion: Since the P-value is greater than α, we fail to reject H0 at the 0.05 level of significance.
There is not enough evidence to conclude that the mean weight of non-top-20 starters is below that of
top-20 teams.
10.81 Population characteristic of interest: μ = average age of brides marrying for the first time in 1990.
Ho: μ = 20.8
Ha: μ > 20.8
α = 0.01
use the t test.
x − 20.8
Test statistic: t =
s
n
23.9 − 20.8 3.10

t= = = 4.84
6.4 0.64
100
P-value = area under the 99 d.f. t curve to the right of 4.84 ≈ 1 − 1 = 0
Conclusion: Since the P-value is less than α, the null hypothesis is rejected at the 0.01 level. The
data supports the conclusion that the mean age of brides marrying for the first time in 1990 is larger
than that in 1970.
10.82 a. Let μ denote the true average salary of a male MBA grad from this Canadian business
school.
Ho: μ = 100,000
Ha: μ > 100,000
α = 0.05 (A value for α was not specified in the problem, so this value was chosen for
illustration.)
x − 100,000
s
n
Computations: n = 258, x = 133,442, s = 131,090
246
133,442 − 100,000 33,442
t= = = 4.0976
131,090 8161.31
258
P-value = area under the 257 d.f. t curve to the right of 4.0976 ≈ 0.0
Since the P-value is smaller than α, Ho is rejected. The sample evidence does support the
conclusion that the average salary of a male MBA grad from this school was in excess of
$100,000 in 1996.
b. Let μ denote the true average salary of a female MBA grad from this Canadian business
school.
Ho: μ = 100,000
Ha: μ > 100,000
α = 0.10
x − 100,000
s
n
105,156 − 100,000 5,156

t= = = 0.7988
98,525 6454.59
233
P-value = area under the 232 d.f. t curve to the right of 0.7988 = 0.2122
Since the P-value is greater than α, Ho is not rejected. The sample evidence does not
support the conclusion that the average salary of a female MBA grad from this school was in
excess of $100,000 in 1996.
c. Because the P-value in b is greater than 0.10, it is also greater than 0.05 and 0.01. Hence,
the same conclusion is reached in b if a significance level of 0.05 or 0.01 is chosen.
10.83 Let π represent the proportion of local residents who oppose hunting on Morro Bay
Ho: π = 0.50
Ha: π > 0.50
α = 0.01
Since nπ = 750(0.50) = 375 ≥ 10, and n(1 − π) = 750(0.5) = 375 ≥ 10, the large sample z test may be
used.
560
n = 750, x = 560, p = = 0.7467
750
247
0.7467 − 0.5 0.2467
z= = = 13.51
0.5(0.5) 0.0183
750
P-value = area under the z curve to the right of 13.51 ≈ 1 − 1 = 0
The data supports the conclusion that the majority of local residents oppose hunting on Morro Bay.
10.84 Let π represent the true population proportion of uninjured occupants.
Ho: π = 0.25
Ha: π > 0.25
α = 0.05
p− 0.25
z=
0.25(0.75)
n
95
n = 319, x = 95, p = = 0.297806
319
0.297806 − 0.25 0.047806

z= = = 1.97
0.25(0.75) 0.024244
.
319
Since the P-value is less than α, Ho is rejected. There is sufficient evidence in this sample to support
the conclusion that the true proportion of uninjured occupants exceeds 0.25.
10.85 Let π denote the true proportion of all cars purchased in this area that were white.
Ho: π = 0.20
Ha: π =/ 0.20
α = 0.05 and/or 0.01. See the conclusion below
n = 400, x = 100, p = 0.25

0.25 − 0.20
z= = 2.50
0.2(0.8)
.
400
P-value = 2(area under the z curve to the right of 2.50) = 2(1 − 0.9938) = 0.0124.
With α = 0.05, the P-value is less than α and the null hypothesis would be rejected. The conclusion
would be that the true proportion of cars sold that are white differs from the national rate of 0.20.
With α = 0.01, the P-value is greater than α and the null hypothesis would not be rejected. Then the
sample does not support the conclusion that the true proportion of cars sold that are white differs from
0.20.
248
10.86 a *
b **
c none
d ***
10.87 Let μ denote the true average time to change (in months)..
Ho: μ = 24
Ha: μ > 24
α = 0.01
x − 24
Test statistic: t =
s
n
35.02 − 24
t= = 3.86
18.94
44
P-value = area under the 43 d.f. t curve to the right of 3.86 ≈ 1 − 1 = 0.
Since the P-value is less than α, Ho is rejected. There is sufficient information in this sample to
support the conclusion that the true average time to change exceeds two years.
10.88 Let π represent the true proportion of all firms who offer stock ownership plans to employees because
of tax-related benefits.
Ho: π = 0.5
Ha: π > 0.5
α = 0.05 (A value for α was not specified in the problem, so this value was chosen.)
p− 0.5
z=
0.5(0.5)
n
54
n = 87, x = 54, p = = 0.62069
87
0.62069 − 0.5 0.12069

z= = = 2.25
0.5(0.5) 0.053606
87
Since the P-value is less than α, Ho is rejected. At level of significance 0.05, the data does support
the conclusion that the proportion of firms who offer stock ownership to employees because of tax-
related benefits exceeds 0.5.
249
10.89 a. Daily caffeine consumption cannot be a negative value. Since the standard deviation is
larger than the mean, this would imply that a sizable portion of a normal curve with this mean
and this standard deviation would extend into the negative values on the number line.
Therefore, it is not plausible that the population distribution of daily caffeine consumption is
normal.
Since the sample size is large (greater than 30) the Central Limit Theorem allows for the
conclusion that the distribution of x is approximately normal even though the population
distribution is not normal. So it is not necessary to assume that the population distribution of
daily caffeine consumption is normal to test hypotheses about the value of population mean
consumption.
b. Let μ denote the population mean daily consumption of caffeine.
Ho : μ = 200
Ha : μ > 200
α = 0.10
x − 200
t=
s
n
n = 47, x = 215, s = 235
x − 200 215 − 200

t= = = 0.44
s 235
n 47
P-value = area under the 46 d.f. t curve to the right of 0.44 = 1 − 0.6700 = 0.33
Since the P-value exceeds the level of significance of 0.10, Ho is not rejected. The data
does not support the conclusion that the population mean daily caffeine consumption
exceeds 200 mg.
10.90 Let π represent the response rate when the distributor is stigmatized by an eye patch.
Ho: π = 0.40
Ha: π > 0.40
α = 0.05
p− 0.40
The test statistic is: z = .
(0.40)(0.60)
n
109
From the data: n = 200, p = = 0.545,
200
0.545 − 0.40
z= = 4.19
(0.40)(0.60)
200
P-value = area under the z curve to the right of 4.19 ≈ 1 − 1 = 0.
Since the P-value is less than the α value of 0.05, Ho is rejected. The data strongly suggests that the
response rate does exceed the rate in the past.
250
10.91 Population characteristic of interest: μ = true average fuel efficiency
Ho: μ = 30
Ha: μ < 30
α = 0.05 (for demonstration purposes)
x − 30
Test statistic: t =
s
n
a normal population distribution. Since the sample size is only 6, we can look at a box plot of the
data. It shows symmetry, indicating that it would not be unreasonable to assume that the population
would be approximately normal. Hence we can proceed with a t-test.
Computations: n = 6, x = 29.33 , s = 1.41
29.33 − 30
t= = − 1.164
1.41
6
P-value = area under the 5 d.f. t curve to the left of -1.164 = 0.15
Conclusion: Since the P-value is greater than α, the null hypothesis is not rejected at the 0.01 level.
The data does not contradict the prior belief that the true average fuel efficiency is at least 30.
10.92 Let μ denote the true average daily revenue (in $) subsequent to the price increase.
Ho: μ = 75
Ha: μ < 75
α = 0.05
x − 75
The test statistic is: t = with d.f. = 19.
s
n
The days are presumed to be representative of all days so we can assume a random sample. It is
likely that the population distribution of weekly sales is normally distributed, so it is reasonable to
proceed with the t test.
From the sample: n = 20, x = 70, s = 4.20,
70 − 75
t= = − 5.32
4.2
20
P-value = area under the 19 d.f. t curve to the left of −5.32 ≈ 0.000.
Since the P-value is less than α, the null hypothesis is rejected. The data does suggest that the true
average daily revenue has decreased from its value prior to the price increase.
251
10.93 Let μ denote the true mean time required to achieve 100o F with the heating equipment of this
manufacturer.
Ho: μ = 15
Ha: μ > 15
α = 0.05
x − 15
The test statistic is: t = .
s
n
From the sample: n = 25, x = 17.5, s = 2.2,
17.5 − 15
t= = 5.68.
2.2
25
P-value =area under the 24 d.f. t curve to the right of 5.68 ≈ 1 − 1 = 0. Because the P-value is
smaller than α , Ho is rejected. The data does cast doubt on the company’s claim that it requires at
most 15 minutes to achieve 100o F.
10.94 a. Let π denote the proportion of voters in a certain state who favor a particular proposed
constitutional amendment.
Ho: π = 0.5
Ha: π > 0.5
α = 0.05
p− 0.5
The test statistic is: z = .
(0.5)(0.5)
n
My sample of random numbers is:
3 1 3 0 9 7 2 5 7 9 0 0 3 4 1 2 7 3 7 9 2 1 5 3 2
8 5 5 7 5 1 3 2 5 1 9 3 3 8 3 8 7 5 5 5 2 4 4 3 4
Let the numbers 0, 1, 2, 3, and 4 represent favoring the proposed constitutional amendment.
n = 50, x = 28, p = 0.56
0.56 − 0.5 0.06

z= = = 0.85
0.5(0.5) 0.070711
50
Since the P-value of 0.198 exceeds the alpha of 0.05, Ho is not rejected. The data does not
support the conclusion that the true proportion of voters who favor a particular proposed
constitutional amendment exceeds 0.5.
252
b. About 0.05(100) = 5 times.
c. In this sample, let the numbers 0, 1, 2, 3, 4, and 5 represent favoring the constitutional
amendment.
9 6 6 4 2 0 9 5 1 9 4 2 6 4 2 6 5 6 1 9 2 3 7 8 8
0 1 5 0 1 5 0 5 0 4 6 9 5 7 3 7 0 0 6 4 3 1 7 1 9
0.62 − 0.5 0.12

n = 50, x = 31, p = 0.62, z= = = 1.697
0.5(0.5) 0.070711
50
Since the P-value of 0.0446 is less than the alpha of 0.05, Ho is rejected. The data does
support the conclusion that the true proportion of voters who favor a particular proposed
constitutional amendment exceeds 0.5. You would expect (and hope) that you would reject a
false null hypothesis more often than you would reject a true null hypothesis.
10.95 P(Type I error) = P(α) = 0

P(Type II error) = P(β) = 0.1 ≤ P(β) ≤ 0.3
253
Chapter 11
11.1 μx = μ1 − μ2 = 30 − 25 = 5
1− x2
σ 12 σ 22 (2)2 (3)2 4 9
σ x1 − x2 = + = + = + = 0.28 = 0.529
n1 n2 40 50 40 50
Since both n1 and n 2 are large, the sampling distribution of x1 − x2 is approximately normal. It is
centered at 5 and the standard deviation is 0.529.
11.2 a. Ho: μ1 − μ 2 = 10 Ha: μ1 − μ 2 > 10
b. Ho: μ1 − μ 2 = −10 Ha: μ1 − μ 2 < −10
11.3 a. The population distributions are approximately normal or the sample size is large and the
two samples are independently selected random samples. Since one of the sample sizes is
only 22 and we don’t have access to the raw data, we must assume that the population
distributions are approximately normal.
b. Let μ1 = mean HRV for all heart attack patients who own dogs.
. Let μ2 = mean HRV for all heart attack patients who do not own dogs.
H 0 : μ1 − μ 2 = 0 H a : μ1 − μ 2 ≠ 0
α = 0.05
( x1 − x2 ) − 0
Test statistic: t=
2 2
s1 s 2
+
n1 n 2
Assumptions have been discussed in part (a)
n1 = 22, x 1 = 873, s1 = 136, n 2 = 80, x 2 = 800, s2 = 134
(873 − 800) − 0
t= = 2.24
(136)2 (134)2
+
22 80
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
df = ⎝ n1 n2 ⎠ =
1134602.622
= 33.08
1 ⎛ s12
⎞
2
1 ⎞ ⎛ s22
2 33658.207 + 637.694
⎜⎜ ⎟⎟ + ⎜⎜ ⎟⎟
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎝ n2 ⎠
So df = 33 (rounded down to an integer)
P-value = 2(the area under the 39 df t curve to the right of 2.24 ≈ 2(0.016) = 0.032.Since the
P-value is less than α, the null hypothesis can be rejected at the 0.05 level of significance.
There is enough evidence to show that mean HRV levels in different in heart attack patients
who own a dog and who do not own a dog.
255
11.4 a. Because the standard deviation is nearly as large as the mean the population distribution will
not fit the 68-95-99.8 rule for normal distributions without the number of hours spent on the
Internet becoming negative; clearly impossible. The population distribution is skewed to the
right.
b. It is appropriate because the samples used are large enough (228 and 306) that the sampling
distribution of the difference of sample means will be approximately normal.
c. Let μ1 = mean number of hours spent online by a typical male teenager

. Let μ2 = mean number of hours spent online by a typical female teenager.
H 0 : μ1 − μ 2 = 0 H a : μ1 − μ 2 > 0
α = .05
( x1 − x2 ) − 0
Test statistic: t=
2 2
s1 s 2
+
n1 n 2
We must be willing to assume that the two samples can be viewed as independently selected
random samples from the two populations of interest and that the samples are large or that
the population distributions are approximately normal. It is stated that the teenagers were
randomly selected and sample sizes are large (228 and 306)
n1 = 228, x 1 = 15.1, s1 = 11.4, n 2 = 306, x 2 = 14.1, s2 = 11.8
(15.1 − 14.1) − 0
t= = .988
(11.4)2 (11.8)2
+
228 306
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
df = ⎝ n1 n2 ⎠ =
1.05
= 505.08
1 ⎛ s12 ⎞
2
1 ⎛ s22 ⎞
2 0.0014 + 0.000679
⎜⎜ ⎟⎟ + ⎜ ⎟
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎜⎝ n2 ⎟⎠
P-value = the area under the 505 df t curve to the right of 0.988 ≈ 0.162.
Since the P-value is greater than α, the null hypothesis cannot be rejected at the 0.05 level of
significance. There is not enough evidence to show that mean number of hours spent online
in a typical week is greater for male teenagers than for female teenagers.
11.5 Let μ1 = mean length of daily commute of a male working adult living in Calgary.
. Let μ2 = mean length of daily commute of a female working adult living in Calgary.
H 0 : μ1 − μ 2 = 0 H a : μ1 − μ 2 ≠ 0
α = 0.05
256
( x1 − x2 ) − 0
Test statistic: t=
2 2
s1 s 2
+
n1 n 2
The sample sizes are large (247 and 253) and it is stated that the samples are independently
selected random samples.
n1 = 247, x 1 = 29.6, s1 = 24.3, n 2 = 253, x 2 = 27.3, s2 = 24.0
(29.6 − 27.3) − 0
t= = 1.06
(24.3)2 (24)2
+
247 253
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
df = ⎝ n1 n2 ⎠ =
21.7839
= 497.36
1 ⎛ s12 ⎞
2
1 ⎛ s22 ⎞
2 0.02323 + 0.02057
⎜ ⎟ + ⎜ ⎟
n1 − 1 ⎜⎝ n1 ⎟⎠ n2 − 1 ⎜⎝ n2 ⎟⎠
P-value = 2(the area under the 497 df t curve to the right of 1.06) ≈ 2(0.143) = 0.2876
Since the P-value is greater than α, the null hypothesis cannot be rejected at the 0.05 level of
significance. There is not enough evidence to show that there is a difference in mean
commute times for male and female working Calgary residents..
11.6 a. We should be willing to assume that the two samples of teens can be viewed as
independently selected random samples. As the sample sizes are large (663 and 413),
the two sample t confidence interval is appropriate.
b. Let μ1 denote the true mean number of calories that a non-fast food eating teen consumes
during the day and μ 2 denote the true mean number of calories that a fast food eating teen
consumes during the day.
n1 = 663, x 1 = 2258, s1 = 1519, n 2 = 413, x 2 = 2637, s2 = 1138.
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
df = ⎝ n1 n2 ⎠ =
9836093.154
= 233.3
1 ⎛ s12 ⎞
2
1 ⎛ s22 ⎞
2 18295.574 + 23865.565
⎜ ⎟ + ⎜ ⎟
n1 − 1 ⎜⎝ n1 ⎟⎠ n2 − 1 ⎜⎝ n2 ⎟⎠
df = 233 so t critical value ≈ 1.97.

Confidence level: not specified, but for demonstration:) = 95%:
(1519)2 (1138)2
(2258 − 2637) ± 1.976 +
663 413
⇒ − 379 ± 1.97(81.338) ⇒ − 379 ± 160.24 ⇒ (−539.24, − 218.76)
257
Based on this sample, we are highly confident that μ1 − μ2 is between −-539.24and -218.76.
Therefore, with 95% confidence, we believe that the mean number of calories that a non-fast
food eating teen consumes in a typical day is between 539.24 and 218.76 calories less than
the mean number consumed by a fast-food eating teen.
11.7 a. H 0 : μ1 = μ2 vs. H a : μ1 < μ2 , where μ1 is the mean payment for all claims not involving
errors and μ 2 is the mean payment for all claims involving errors.
b. With sample sizes so large (515 and 889) the df will be large, and the bottom row of the t
table can be used. With a P-value of 0.004, and a lower-tailed test, the value of the test
statistic must have been between 2.58 and 3.09. t = 2.65 is the best answer.
11.8 Let μ1 = mean GPA of students at University of Central Florida that are employed
. Let μ2 = mean GPA of students at University of Central Florida that are not employed.
H 0 : μ1 − μ 2 = 0 H a : μ1 − μ 2 < 0
α = 0.05 (not specified, but used for demonstration purposes)
( x1 − x2 ) − 0
Test statistic: t=
2 2
s1 s 2
+
n1 n 2
The sample sizes are large (184 and 114) and it is stated that the samples are selected at random.
n1 = 184, x 1 = 3.12, s1 = .485, n 2 = 114, x 2 = 3.23, s2 = .524
(3.12 − 3.23) − 0
t= = −1.81
(.485)2 (.524)2
+
184 114
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
⎝ n1 n2 ⎠ 1.37(10−5 )
df = = = 227.295
1 ⎛ s12 ⎞
2
1 ⎛ s22 ⎞
2
8.93(10−9 ) + 5.13(10−8 )
⎜⎜ ⎟⎟ + ⎜ ⎟
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎜⎝ n2 ⎟⎠
P-value = the area under the 227 df t curve to the left of -1.81 ≈ 0.036
Since the P-value is less than α, the null hypothesis can be rejected at the 0.05 level of
significance. There is enough evidence to show that for students at the University of
Central Florida, those who are not employed have a higher mean GPA than those who
are employed.
258
11.9 a. Let μ1 denote the true mean breaking force in a dry medium at 37 degrees and μ 2 denote
the true mean breaking force in a wet medium at 37 degrees.
Although the distribution of the sample taken from both conditions are skewed, there are no
outliers and it would not be unreasonable to assume they both came from approximately
normal populations. We would have to assume the samples were independent and taken at
random.
37 dry
37 wet
300 320 340 360 380 400

Force
n1 =6 , x 1 = 311.6, s1 = 18.4, n 2 = 6, x 2 = 355.6, s2 = 27.3.
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
df = ⎝ n1 n2 ⎠ =
32631.424
= 8.765
1 ⎛ s12 ⎞
2
1 ⎞ ⎛ s22
2 636.79 + 3085.873
⎜⎜ ⎟⎟ + ⎜⎜ ⎟⎟
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎝ n2 ⎠
df = 8 so t critical value =1.86.
Confidence level: 90%:
(18.4)2 (27.3)2
(311.6 − 355.6) ± 1.86 +
6 6
⇒ − 44 ± 1.86(13.44) ⇒ − 44 ± 25.0 ⇒ ( −69, − 19)
Based on this sample, we are highly confident that μ1 − μ2 is between -69 and -19.
Therefore, with 90% confidence, we believe that the mean breaking force required to break a
cement bond is between 19 and 69 Newtons greater in a wet medium than a dry medium
when the temperature is 37 degrees.
b. Let μ1 denote the true mean breaking force in a dry medium at 37 degrees and μ2 denote
the true mean breaking force in a dry medium at 22 degrees.
H 0 : μ1 − μ 2 = 100 H a : μ1 − μ 2 > 100
α = 0.10
( x1 − x2 ) − 100
Test statistic: t=
2 2
s1 s 2
+
n1 n 2
259
37 dry
22 dry
100 150 200 250 300 350

Force
Although the distribution of the sample taken from both conditions are skewed, there are no
outliers and it would not be unreasonable to assume they both came from approximately
normal populations. We would have to assume the samples were independent and taken at
random.
n1 = 6, x 1 = 311.6, s1 = 18.4, n 2 = 6, x 2 = 157.5, s2 = 44.3
(311.6 − 157.5) − 100

t= = 2.76
(18.4)2 (44.3)2
+
6 6
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
df = ⎝ n1 n2 ⎠ =
147081.20
= 6.75
1 ⎛ s12 ⎞
2
1 ⎛ s22 ⎞
2 636.64 + 21396.27
⎜⎜ ⎟⎟ + ⎜ ⎟
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎜⎝ n2 ⎟⎠
P-value = the area under the 6 df t curve to the right of 2.76 ≈ 0.016
significance. There is enough evidence to show that the mean breaking force in a dry
medium at the higher temperature is greater than the mean breaking force at the lower
temperature by more than 100 N.
11.10 Let μ1 denote the true mean percentage of the time that the genetically altered voles spent with the
previous partner and μ2 denote the true mean percentage of the time that the unaltered
voles spent with the previous partner.
H 0 : μ1 − μ 2 = 0 H a : μ1 − μ 2 > 0
α = 0.05
260
( x1 − x2 ) − 0
Test statistic: t=
2 2
s1 s 2
+
n1 n 2
Altered
Unaltered
0 20 40 60 80 100
% of Time
The sample data for both groups is slightly skewed but there are no outliers and so it is not
unreasonable to assume that the data comes from populations that are approximately normal.
If we assume that the voles were selected as independent random samples, the two sample t-test for
the difference of means is the appropriate test.
n1 = 11, x 1 = 82.6, s1 = 13.1, n 2 = 20, x 2 = 56.8, s2 = 29.7
(82.6 − 56.8) − 0
t= = 3.34
(13.1)2 (29.7)2
+
11 20
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
df = ⎝ n1 n2 ⎠ =
3564.63
= 28.14
1 ⎛ s12 ⎞
2
1 ⎛ s22 ⎞
2 24.336 + 102.358
⎜⎜ ⎟⎟ + ⎜ ⎟
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎜⎝ n2 ⎟⎠
significance. There is enough evidence to show that the mean percentage of time spent with
the previous partner is significantly greater for genetically altered voles than for voles that did
not have the gene introduced.
11.11 a. Let μ1 be the true mean percentage of time male monkeys spend playing with a police car
and
μ2 be the true mean percentage of the time female monkeys spend playing with a police car.
Ho: μ1 − μ 2 = 0 Ha: μ1 − μ 2 > 0
α = 0.05
( x1 − x2 ) − 0
Test statistic: t=
2 2
s1 s 2
+
n1 n 2
261
The sample sizes for the two groups are large (both 44) and we can assume that the two
samples are representative of the populations of male and female monkeys so we can regard
them as random samples.
n1 = 44, x 1 = 18, s1 = 5, n 2 = 44, x 2 = 8, s2 = 4
(18 − 8) − 0
t= = 10.4
(5)2 (4)2
+
44 44
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
df = ⎝ n1 n2 ⎠ =
.8678
= 82.01
1 ⎛ s12 ⎞
2
1 ⎛ s22 ⎞
2 0.0075 + .0031
⎜⎜ ⎟⎟ + ⎜ ⎟
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎜⎝ n2 ⎟⎠
Since the P-value is much less than α, the null hypothesis of no difference is rejected. There
is very strong evidence to indicate that that the mean percentage of the time spent playing
with the police car is greater for male monkeys than for female monkeys.
b. Let μ1 be the true mean percentage of time female monkeys spend playing with a doll and
μ2 be the true mean percentage of the time male monkeys spend playing with a doll.
Ho: μ1 − μ 2 = 0 Ha: μ1 − μ 2 > 0
α = 0.05
( x1 − x2 ) − 0
Test statistic: t=
2 2
s1 s 2
+
n1 n 2
n1 = 44, x 1 = 20, s1 = 4, n 2 = 44, x 2 = 9, s2 = 2
(20 − 9) − 0
t= = 16.3
(4)2 (2)2
+
44 44
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
df = ⎝ n1 n2 ⎠ =
.2069
= 62.88
1 ⎛ s12 ⎞
2
1 ⎛ s22 ⎞
2 .0031 + .00019
⎜⎜ ⎟⎟ + ⎜ ⎟
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎜⎝ n2 ⎟⎠
262
Since the P-value is much less than α, the null hypothesis can be rejected. There is very
strong evidence to indicate that that the mean percentage of the time spent playing with the
doll is greater for female monkeys than for male monkeys.
c. Let μ1 be the true mean percentage of time female monkeys spend playing with a furry dog
and μ2 be the true mean percentage of the time male monkeys spend playing with a furry
dog.
Ho: μ1 − μ 2 = 0 Ha: μ1 − μ 2 ≠ 0
α = 0.05
( x1 − x2 ) − 0
Test statistic: t=
2 2
s1 s 2
+
n1 n 2
n1 = 44, x 1 = 20, s1 = 5, n 2 = 44, x 2 = 25, s2 = 5
(20 − 25) − 0
t= = −4.69
(5)2 (5)2
+
44 44
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
df = ⎝ n1 n2 ⎠ =
1.29
= 85.96
1 ⎛ s12 ⎞
2
1 ⎞ ⎛ s22
2 .0075 + .0075
⎜⎜ ⎟⎟ + ⎜⎜ ⎟⎟
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎝ n2 ⎠
P-value = 2(the area under the 85 df t curve to the left of -4.69) ≈ 2(0.00) ≈ 0.00.
Since the P-value is much less than α, the null hypothesis of no difference is rejected. There
is very strong evidence to indicate that that the mean percentage of the time spent playing
with the furry dog differs between male and female monkeys.
d. Although it appears that the male monkeys spent time playing with the “boy” toys and the
female monkeys spent more time playing with the “girl” toys, there was a difference in the
“neutral” toy as well. Any difference in the time spent playing with these toys may have
nothing to do with the gender choices, males monkeys may like a certain color better, or
shiny surfaces or some other underlying factor. This is an observational study and as such,
causation cannot be shown.
e. One of the conditions of a two-sample t test is that the two samples are independent. In this
example, if the mean percentage of the time spent playing with the police car increased, the
mean percentage of the time spent playing with the doll would have to decrease. Therefore
one affects the other, they are not independent and this violates one of the conditions of the
test.
263
11.12 a. Let μ1 denote the true average amount by which male teenage drivers exceed the speed
limit and μ 2 denote the true average amount by which female teenage drivers exceed the
speed limit.
H 0 : μ1 − μ 2 = 0 H a : μ1 − μ 2 > 0
α = 0.01
( x1 − x2 ) − 0
Test statistic: t=
2 2
s1 s 2
+
n1 n 2
Male
Female
0.0 0.5 1.0 1.5 2.0 2.5 3.0

Excess Speed
The lack of outliers in the graph of the sample data suggests that it would not be
unreasonable to assume that both samples come from populations that are approximately
normal. We can assume that the two samples are representative of the populations of male
and female teenage drivers so we can regard them as independent random samples.
n1 = 10, x 1 = 1.46, s1 = 0.74, n 2 = 10, x 2 = 0.64, s2 = 0.455
(1.46 − 0.64) − 0
t= = 2.99
(0.74)2 (0.455)2
+
10 10
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
df = ⎝ n1 n2 ⎠ =
.00569
= 14.94
1 ⎛ s12 ⎞
2
1 ⎛ s22 ⎞
2 .000333 + .0000476
⎜ ⎟ + ⎜ ⎟
n1 − 1 ⎜⎝ n1 ⎟⎠ n2 − 1 ⎜⎝ n2 ⎟⎠
Since the P-value is much less than α, the null hypothesis can be rejected. There is very
strong evidence to indicate that that, on average, male teenage drivers exceed the speed
limit by more than do female teenage drivers.
264
b. For all tests, the sample sizes are large and the samples are considered representative of the
population so the conditions of each test is met. Because the sample standard deviations
are all the same, it is not unreasonable to assume that all the population variances are equal.
Using Minitab to perform the Pooled t Test:
i. H 0 : μ1 − μ 2 = 0 vs H a : μ1 − μ 2 > 0 where μ1 is the mean number of miles per

hour over the speed limit for male drivers with male passengers and μ 2 is the mean number
of miles per hour over the speed limit for male drivers with female passengers.
Two-Sample T-Test and CI
Sample N Mean StDev SE Mean
1 40 5.200 0.800 0.13
2 40 0.300 0.800 0.13
Difference = mu (1) - mu (2)
Estimate for difference: 4.90000
T-Test of difference = 0 (vs >): T-Value = 27.39 P-Value = 0.000 DF = 78
Both use Pooled StDev = 0.8000
Since the P-value is much less than any reasonable α, the null hypothesis can be rejected.
There is very strong evidence to indicate that that the average number of miles per hour over
the speed limit is greater for male drivers with male passengers than it is for male drivers with
female passengers.
ii. H 0 : μ1 − μ 2 = 0 vs H a : μ1 − μ 2 > 0 where μ1 is the mean number of miles per

hour over the speed limit for female drivers with male passengers and μ2 is the mean number
of miles per hour over the speed limit for female drivers with female passengers.
1 40 2.300 0.800 0.13
2 40 0.600 0.800 0.13
Estimate for difference: 1.70000
T-Test of difference = 0 (vs >): T-Value = 9.50 P-Value = 0.000 DF = 78
the speed limit is greater for female drivers with male passengers than it is for female drivers
with female passengers.
iii. H 0 : μ1 − μ 2 = 0 vs H a : μ1 − μ 2 < 0 where μ1 is the mean number of miles per

hour over the speed limit for male drivers with female passengers and μ2 is the mean number
of miles per hour over the speed limit for female drivers with male passengers.
1 40 0.300 0.800 0.13
2 40 2.300 0.800 0.13
Estimate for difference: -2.00000
T-Test of difference = 0 (vs <): T-Value = -11.18 P-Value = 0.000 DF = 78
the speed limit is smaller for female drivers with male passengers than it is for female drivers
with male passengers.
c. It appears that it is the gender of the passenger that has the greater effect than the gender of
the driver on the teen drivers speed. Although the amount by which the speed limit was
exceeded is higher for a male teen driver than a female teen driver, if they have a passenger,
a male passenger is associated with a greater average number of miles per hour over the
speed limit than a female passenger regardless of the gender of the driver.
265
11.13 μ1 be the mean Wechsler Memory Scale score taken by the Ginko group
. μ2 be the mean Wechsler Memory Scale score taken by the control group
Ho: μ1 − μ 2 = 0 Ha: μ1 − μ 2 > 0
α = 0.05
( x1 − x2 ) − 0
Test statistic: t=
2 2
s1 s 2
+
n1 n 2
Assumptions: The population distributions are approximately normal or the sample size is large
(generally ≥ 30) and the two samples are independently selected random samples.
n1 = 104, x 1 = 5.6, s1 = 0.6, n 2 = 115, x 2 = 5.5, s2 = 0.6
(5.6 − 5.5) − 0 0.1

t= = = 1.23
2
(0.6) (0.6) 2 0.0812
+
104 115
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
⎝ n1 n2 ⎠ (0.0035 + 0.0031)2
df = 2 2
= = 214.34
1 ⎛ s12 ⎞ 1 ⎛ s22 ⎞ (0.0035)2 (0.0031)2
⎜⎜ ⎟⎟ + ⎜ ⎟ +
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎜⎝ n2 ⎟⎠ 103 114
Since the P-value is greater than α, the null hypothesis is not rejected at the 0.05 level of significance.
There is not enough evidence to show that taking 40 mg ginkgo three times a day is effective in
increasing mean performance on the Wechsler Memory Scale.
11.14 a. To use a 2 sample t- test, one of the requirements is the samples are either large (generally ≥
30) or that the population distributions are approximately normally distributed. In this case
both sample sizes are small (10) and we do not any information about the population
distributions.
b. No the sample sizes are large enough that we can, with the use of the Central Limit Theorem,
assume that the sampling distribution of x1 − x 2 is approximately normal.
c. μ1 be the mean fumonisin level for partially degermed corn meal

. μ2 be the mean fumonisin level for partially degermed corn meal
Ho: μ1 − μ 2 = 0 Ha: μ1 − μ 2 ≠ 0
α = 0.01
( x1 − x2 ) − 0
Test statistic: t=
2 2
s1 s 2
+
n1 n 2
266
Assumptions: The population distributions are approximately normal or the sample size is
large (generally ≥ 30) and the two samples are independently selected random samples.
n1 = 50, x 1 = 0.59, s1 = 1.01, n 2 = 50, x 2 = 1.21, s2 = 1.71
(0.59 − 1.21) − 0 −0.62

t= = = −2.207
(1.01)2 (1.71)2 0.2809
+
50 50
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
⎝ n1 n2 ⎠ (0.0204 + 0.0585)2
df = 2 2
= = 79.47
1 ⎛ s12 ⎞ 1 ⎛ s22 ⎞ (0.0204)2 (0.0585)2
⎜⎜ ⎟⎟ + ⎜ ⎟ +
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎜⎝ n2 ⎟⎠ 49 49
P-value = 2(the area under the 79 df t curve to the left of -2.207) ≈ 0.030
Since the P-value is greater than α, the null hypothesis is not rejected at the 0.01 level of
significance. There is not enough evidence to show that there is a difference in mean
fumonisin level for the two types of corn meal.
11.15 a. Let μ1 be the true mean “appropriateness” score assigned to wearing a hat in a class by the
population of students and μ2 be the corresponding score for faculty.
Ho: μ1 − μ 2 = 0 Ha: μ1 − μ 2 ≠ 0
α = 0.05 (A value for α is not specified in the problem. We use α=0.05 for illustration.)
( x1 − x2 ) − 0
Test statistic: t=
2 2
s1 s 2
+
n1 n 2
Assumptions: The sample sizes for the two groups are large (say, greater than 30 for each)
and the two samples are independently selected random samples.
n1 = 173, x 1 = 2.80, s1 = 1.0, n 2 = 98, x 2 = 3.63, s2 = 1.0
(2.80 − 3.63) − 0 −0.83

t= = = −6.5649
2
(1.0) (1.0) 2 0.1264
+
173 98
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
⎝ n1 n2 ⎠ (0.00578 + 0.01020)2
df = 2 2
= = 201.5
1 ⎛ s12 ⎞ 1 ⎛ s22 ⎞ (0.00578)2 (0.01020)2
⎜⎜ ⎟⎟ + ⎜ ⎟ +
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎜⎝ n2 ⎟⎠ 172 97
267
P-value = 2 times the area under the 201 df t curve to the left of –6.5649 ≈ 0.0000.
Since the P-value is much less than α, the null hypothesis of no difference is rejected. The
data do provide very strong evidence to indicate that there is a difference in the mean
appropriateness scores between students and faculty for wearing hats in the class room.
The mean appropriateness score for students is significantly smaller than that for faculty.
b. Let μ1 be the true mean “appropriateness” score assigned to addressing an instructor by his
or her first name by the population of students and μ2 be the corresponding score for faculty.
Ho: μ1 − μ 2 2 = 0 Ha: μ1 − μ 2 > 0
( x1 − x2 ) − 0
Test statistic: t=
2 2
s1 s 2
+
n1 n 2
n1 = 173, x 1 = 2.90, s1 = 1.0, n 2 = 98, x 2 = 2.11, s2 = 1.0
(2.90 − 2.11) − 0 0.79

t= = = 6.2485
2
(1.0) (1.0) 2 0.1264
+
173 98
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
⎝ n1 n2 ⎠ (0.00578 + 0.01020)2
df = 2 2
= = 201.5
1 ⎛ s12 ⎞ 1 ⎛ s22 ⎞ (0.00578)2 (0.01020)2
⎜⎜ ⎟⎟ + ⎜ ⎟ +
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎜⎝ n2 ⎟⎠ 172 97
Since the P-value is much less than α, the null hypothesis of no difference is rejected. The
data do provide very strong evidence to indicate that the mean appropriateness score for
addressing the instructor by his or her first name is higher for students than for faculty.
c. Let μ1 be the true mean “appropriateness” score assigned to talking on a cell phone
during class by the population of students and μ2 be the corresponding score for faculty.
Ho: μ1 − μ 2 = 0 Ha: μ1 − μ 2 ≠ 0
( x1 − x2 ) − 0
Test statistic: t=
2 2
s1 s 2
+
n1 n 2
268
n1 = 173, x 1 = 1.11, s1 = 1.0, n 2 = 98, x 2 = 1.10, s2 = 1.0
(1.11 − 1.10) − 0 −0.01

t= = = 0.0791
(1.0)2 (1.0)2 0.1264
+
173 98
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
⎝ n1 n2 ⎠ (0.00578 + 0.01020)2
df = 2 2
= = 201.5
1 ⎛ s12 ⎞ 1 ⎛ s22 ⎞ (0.00578)2 (0.01020)2
⎜⎜ ⎟⎟ + ⎜ ⎟ +
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎜⎝ n2 ⎟⎠ 172 97
P-value = 2 times the area under the 201 df t curve to the right of 0.0791 ≈ 0.9370.
Since the P-value is not less than α, the null hypothesis of no difference cannot be rejected.
The data do not provide evidence to indicate that there is a difference in the mean
appropriateness scores between students and faculty for talking on cell phones in class. The
result does not imply that students and faculty consider it acceptable to talk on a cell phone
during class. It simply says that data do not provide enough evidence to claim a difference
exists.
11.16 a. Let μ1 be the true mean stream gradient (%) for the population of sites with tailed frogs
and μ2 be the corresponding mean for sites without tailed frogs.
Ho: μ1 − μ 2 = 0 Ha: μ1 − μ 2 ≠ 0
α = 0.01
( x1 − x2 ) − 0
Test statistic: t=
2 2
s1 s 2
+
n1 n 2
Assumptions: The distribution of stream gradients is approximately normal for both types of
sites and the two samples are independently selected random samples.
n1 = 18, x 1 = 9.1, s1 = 6.0, n 2 = 31, x 2 = 5.9, s2 = 6.29
(9.1 − 5.9) − 0 3.2

t= = = 1.7679
2
(6.00) (6.29) 2 1.8100
+
18 31
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
⎝ n1 n2 ⎠ (2.00 + 1.2763)2
df = 2 2
= = 37.07
1 ⎛ s12 ⎞ 1 ⎛ s22 ⎞ (2.00)2 (1.2763)2
⎜ ⎟ + ⎜ ⎟ +
n1 − 1 ⎜⎝ n1 ⎟⎠ n2 − 1 ⎜⎝ n2 ⎟⎠ 17 30
269
Since the P-value is greater than α, the null hypothesis of no difference cannot be rejected.
The data do not provide sufficient evidence to suggest a difference between the mean stream
gradients for sites with tailed frogs and sites without tailed frogs.
b. Let μ1 be the true mean water temperature for sites with tailed frogs and μ2 be the
corresponding mean for sites without tailed frogs.
Ho: μ1 − μ 2 = 0 Ha: μ1 − μ 2 ≠ 0
α = 0.01
( x1 − x2 ) − 0
Test statistic: t=
2 2
s1 s 2
+
n1 n 2
Assumptions: The distribution of stream gradients is approximately normal for both types of
sites and the two samples are independently selected random samples.
n1 = 18, x 1 = 12.2, s1 = 1.71, n 2 = 31, x 2 = 12.8, s2 = 1.33
(12.2 − 12.8) − 0 −0.6

t= = = −1.2806
(1.71)2 (1.33)2 0.4685
+
18 31
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
⎝ n1 n2 ⎠ (0.1625 + 0.0571)2
df = 2 2
= = 29.01
1 ⎛ s12 ⎞ 1 ⎛ s22 ⎞ (0.1625)2 (0.0571)2
⎜⎜ ⎟⎟ + ⎜ ⎟ +
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎜⎝ n2 ⎟⎠ 17 30
P-value = 2 times the area under the 29 df t curve to the left of -1.2806 ≈ 0.2105.
Since the P-value is greater than α, the null hypothesis of no difference cannot be rejected.
The data do not provide sufficient evidence to conclude that the mean water temperatures for
the two types of sites (with and without tailed frogs) are different.
c. Let μ1 be the true mean stream depth for the population of sites with tailed frogs and μ2
be the corresponding mean for sites without tailed frogs.
Ho: μ1 − μ 2 = 0 Ha: μ1 − μ 2 ≠ 0
α = 0.01
( x1 − x2 ) − 0
Test statistic: t=
2 2
s1 s 2
+
n1 n 2
270
Assumptions: The sample sizes for each group is large (greater than or equal to 30) and the
two samples are independently selected random samples. Since the sample sizes for the
two samples are 82 and 267 respectively, it is quite reasonable to use the independent
samples t test for comparing the mean depths for the two types of sites.
n1 = 82, x 1 = 5.32, s1 = 2.27, n 2 = 267, x 2 = 8.46, s2 = 5.95
(5.32 − 8.46) − 0 −3.14

t= = = −7.1028
2
(2.27) (5.95) 2 0.4421
+
82 267
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
⎝ n1 n2 ⎠ (0.06284 + 0.13259)2
df = 2 2
= = 332.6
1 ⎛ s12 ⎞ 1 ⎛ s22 ⎞ (0.06284)2 (0.13259)2
⎜⎜ ⎟⎟ + ⎜ ⎟ +
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎜⎝ n2 ⎟⎠ 81 266
Since the P-value is much smaller than α, the null hypothesis of no difference is rejected.
The data provide very strong evidence to conclude that there indeed is a difference between
the mean stream depths of sites with tailed frogs and sites without tailed frogs.
11.17 Let μ1 denote the true mean “intention to take science courses” score for male students and μ2 be the
corresponding score for female students. Then μ1 − μ2 denotes the difference between the means of
the intention scores for males and females.
n1 = 203, x1 = 3.42, s1 = 1.49, n2 = 224, x2 = 2.42, s2 = 1.35
s12 (1.49)2 s22 (1.35)2

V1 = = = 0.01094 V2 = = = 0.00814
n1 203 n2 224
(V1 + V2 )2 (0.01094 + 0.00814)2

df = = = 409.2
V12 V2 (0.01094)2 (0.00814)2
+ 2 +
n1 − 1 n2 − 1 202 223
Use df = 409. The t critical value is 2.588.
The 95% confidence interval for μ1 − μ2 based on this sample is
(3.42 − 2.42) ± 2.588 0.01094 + 0.00814 ⇒ 1 ± 2.588(0.13811) ⇒ (0.6426,1.3574) .
Observe that the interval does not include 0, and so 0 is not one of the plausible values of μ1 − μ2. As
a matter of fact, the plausible values are in the interval from 0.6426 to 1.3574. The data provide
sufficient evidence to conclude that the mean “intention to take science courses” scores for male
students is greater than that for female students.
271
11.18 Because the sample sizes are small ( n1 = 10 and n2 = 10 ) it is important to check the assumption of
normality of the two populations. Boxplots of the two samples, shown in the figure below, suggest
that the population distributions of weights are approximately symmetric and no outliers appear to be
present. So, it might not be unreasonable to assume approximate normality. Also, it is reasonable to
assume that the two samples are independently selected random samples. Therefore it is reasonable
to use the independent samples t test to compare the mean weights of fish for 1995 and 1996.
1996
Year
1995
550 650 750 850 950 1050

Weight
11.19 a. Let μ1 denote the true mean hardness for chicken chilled 0 hours before cooking and μ2
the true mean hardness for chicken chilled 2 hours before cooking.
Ho: μ1 − μ 2 = 0 Ha: μ1 − μ 2 ≠ 0
α = 0.05
( x1 − x2 ) − 0
Test statistic: t=
2 2
s1 s 2
+
n1 n 2
two samples are independently selected random samples.
n1 = 36, x 1 = 7.52, s1 = 0.96, n 2 = 36, x 2 = 6.55, s2 = 1.74
(7.52 − 6.55) − 0 0.97

t= = = 2.93
(0.96)2 (1.74)2 0.33121
+
36 36
272
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
⎝ n1 n2 ⎠ (0.0256 + 0.0841)2
df = 2 2
= = 54.5
1 ⎛ s12 ⎞ 1 ⎛ s22 ⎞ (0.0256)2 (0.0841)2
⎜⎜ ⎟⎟ + ⎜ ⎟ +
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎜⎝ n2 ⎟⎠ 35 35
P-value = 2(area under the 54 df t curve to the right of 2.93) = 2(1 − 0.9975) =
2(0.00249) = 0.00498.
Since the P-value is less than α, the null hypothesis is rejected. At level of significance 0.05,
there is sufficient evidence to conclude that there is a difference in mean hardness of chicken
chilled 0 hours before cooking and chicken chilled 2 hours before cooking.
b. Let μ1 denote the true mean hardness for chicken chilled 8 hours before cooking and μ2
the true mean hardness for chicken chilled 24 hours before cooking.
Ho: μ1 − μ 2 = 0 Ha: μ1 − μ 2 ≠ 0
α = 0.05
( x1 − x2 ) − 0
Test statistic: t=
2 2
s1 s 2
+
n1 n 2
n1 = 36, x 1 = 5.70, s1 = 1.32, n 2 = 36, x 2 = 5.65, s2 = 1.50
(5.70 − 5.65) − 0 0.05

t= = = 0.15
2
(1.32) (1.50) 2 0.333017
+
36 36
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
⎝ n1 n2 ⎠ (0.0484 + 0.0625)2
df = 2 2
= = 68.9
1 ⎛ s12 ⎞ 1 ⎛ s22 ⎞ (0.0484)2 (0.0625)2
⎜⎜ ⎟⎟ + ⎜ ⎟ +
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎜⎝ n2 ⎟⎠ 35 35
P-value = 2(area under the 68 df t curve to the right of 0.15) = 2(1 − 0.5595)
= 2(0.44055) = 0.8811.
Since the P-value exceeds α, the null hypothesis is not rejected. At level of significance 0.05,
there is not sufficient evidence to conclude that there is a difference in mean hardness of
chicken chilled 8 hours before cooking and chicken chilled 24 hours before cooking.
c. Let μ1 denote the true mean hardness for chicken chilled 2 hours before cooking and μ2 the
true mean hardness for chicken chilled 8 hours before cooking.
n1 = 36, x 1 = 6.55, s1 = 1.74, n 2 = 36, x 2 = 5.70, s2 = 1.32

273
(1.74)2 (1.32)2
(6.55 − 5.70) ± 1.669 +
36 36
⇒ .85 ± 1.669(.364005) ⇒ .85 ± .6075 ⇒ (.242, 1.458)
Based on this sample, we believe that the mean hardness for chicken chilled for 2 hours
before cooking is larger than the mean hardness for chicken chilled 8 hours before cooking.
The difference may be as small as 0.242, or may be as large as 1.458.
11.20 Let μ1 denote the true mean RMA for fraternity members and μ2 the true mean RMA for non-fraternity
members.
Ho: μ1 − μ 2 = 0 Ha: μ1 − μ 2 < 0
α = 0.05
( x1 − x2 ) − 0
Test statistic: t=
2 2
s1 s 2
+
n1 n 2
Assumptions: The sample size for each group is large (greater than or equal to 30) and the two
samples are independently selected random samples.
n1 = 178, x 1 = 25.63, s1 = 6.16, n 2 = 155, x 2 = 27.40, s2 = 5.51
(25.63 − 27.4) − 0 −1.77

t= = = −2.77
2
(6.16) (5.51) 2 .639569
+
178 155
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
⎝ n1 n2 ⎠ (0.2132 + 0.1959)2
df = 2 2
= = 330.8
1 ⎛ s12 ⎞ 1 ⎛ s22 ⎞ (0.2132)2 (0.1959)2
⎜⎜ ⎟⎟ + ⎜ ⎟ +
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎜⎝ n2 ⎟⎠ 177 154
P-value = area under the 330 df t curve to the left of −2.77 = 0.00298.
Since the P-value is less than α, the null hypothesis is rejected. The data supports the conclusion
that the true mean RMA score for fraternity members is lower than that of non-fraternity members.
11.21 Let μ1 denote the true mean alkalinity for upstream locations and μ2 the true mean alkalinity for
downstream locations.
Ho: μ1 − μ 2 = -50 Ha: μ1 − μ 2 < -50
α = 0.05
274
( x1 − x2 ) − ( −50)
Test statistic: t=
2 2
s1 s 2
+
n1 n 2
Assumptions: The distribution of alkalinity is approximately normal for both types of sites (upstream
and downstream) and the two samples are independently selected random samples.
n1 = 24, x 1 = 75.9, s1 = 1.83, n 2 = 24, x 2 = 183.6, s2 = 1.70
(75.9 − 183.6) − ( −50) −57.7

t= = = 113.17
2
(1.83) (1.70) 2 0.50986
+
24 24
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
⎝ n1 n2 ⎠ (0.1395 + 0.1204)2
df = 2 2
= = 45.75
1 ⎛ s12 ⎞ 1 ⎛ s22 ⎞ (0.1395)2 (0.1204)2
⎜⎜ ⎟⎟ + ⎜ ⎟ +
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎜⎝ n2 ⎟⎠ 23 23

P-value = area under the 45 df t curve to the right of 113 ≈ 0.
that the true mean alkalinity score for downstream sites is more than 50 units higher than that for
upstream sites.
11.22 Let μ1 denote the true mean cholesterol level of people who have attempted suicide and μ2 the true
mean cholesterol level of people who have not attempted suicide.
Ho: μ1 − μ 2 = 0 Ha: μ1 − μ 2 < 0
α = 0.05
( x1 − x2 ) − 0
Test statistic: t=
2 2
s1 s 2
+
n1 n 2
n1 = 331, x 1 = 198, s1 = 20, n 2 = 331, x 2 = 217, s2 = 24
(198 − 217) −19

t= = = −11.06
2
(20) (24) 2 1.717161
+
331 331
275
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
⎝ n1 n2 ⎠ (1.2085 + 1.7402)2
df = 2 2
= = 639.2
1 ⎛ s12 ⎞ 1 ⎛ s22 ⎞ (1.2085)2 (1.7402)2
⎜⎜ ⎟⎟ + ⎜ ⎟ +
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎜⎝ n2 ⎟⎠ 330 330
P-value = area under the 639 df t curve to the left of −11.06 ≈ 0.
that the true mean cholesterol level of people who attempt suicide is less than the mean cholesterol
level of people who do not attempt suicide.
11.23 Let μ1 denote the mean frequency of alcohol use for those that rush a sorority and μ2 denote the
mean frequency of alcohol use for those that do not rush a sorority.
Ho: μ1 − μ 2 = 0 Ha: μ1 − μ 2 > 0
α = 0.01
( x1 − x2 ) − 0
Test statistic: t=
2 2
s1 s 2
+
n1 n 2
n1 = 54, x 1 = 2.72, s1 = 0.86, n 2 = 51, x 2 = 2.11, s2 = 1.02
(2.72 − 2.11) − 0 0.61

t= = = 3.30
(0.86)2 (1.02)2 0.184652
+
54 51
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
⎝ n1 n2 ⎠ (0.0137 + 0.0204)2
df = 2 2
= = 98.002
1 ⎛ s12 ⎞ 1 ⎛ s22 ⎞ (0.0137)2 (0.0204)2
⎜⎜ ⎟⎟ + ⎜ ⎟ +
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎜⎝ n2 ⎟⎠ 53 50
P-value = area under the 98 df t curve to the right of 3.30 = 1 − 0.9993 = 0.0007
that the true mean frequency of alcohol use is larger for those that rushed a sorority than for those
who did not rush a sorority.
276
11.24 Let μ1 denote the mean force after impact for advanced players and μ2 denote the
mean force after impact for intermediate players.
Ho: μ1 − μ 2 = 0 Ha: μ1 − μ 2 > 0
α = 0.01
( x1 − x2 ) − 0
Test statistic: t=
2 2
s1 s 2
+
n1 n 2
Assumptions: For the two sample t-test to be appropriate, we must be willing to assume that the two
samples can be viewed as independently selected random samples from the population of advanced
and intermediate tennis players. Since the samples are both small, it is necessary to assume that the
force distribution is approximately normal for each of these populations. Boxplots constructed using
the sample data show symmetric distributions consistent with the assumption of normal population
distributions.
n1 = 6, x 1 = 40.3, s1 = 11.3, n 2 = 8, x 2 = 21.4, s2 = 8.3
(40.3 − 21.4) − 0 18.9

t= = = 3.46
2
(11.3) (8.3) 2 5.467
+
6 8
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
⎝ n1 n2 ⎠ (21.282 + 8.611)2
df = 2 2
= = 8.8321
1 ⎛ s12 ⎞ 1 ⎛ s22 ⎞ (21.282)2 (8.611)2
⎜⎜ ⎟⎟ + ⎜ ⎟ +
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎜⎝ n2 ⎟⎠ 5 7
P-value = area under the 8 df t curve to the right of 3.50 = 0.004
Since the P-value is less than α, the null hypothesis is rejected. There is sufficient evidence to
conclude that the mean force after impact is greater for advanced tennis players than it is for
intermediate players.
11.25 Let μ1 denote the mean half-life of vitamin D in plasma for people on a normal diet. Let μ2 denote
the mean half-life of vitamin D in plasma for people on a high-fiber diet. Let μ1 − μ2 denote the true
difference in mean half-life of vitamin D in plasma for people in these two groups (normal minus
high fiber).
Ho: μ1 − μ 2 = 0 Ha: μ1 − μ 2 > 0
α = 0.01
( x1 − x2 ) − 0
Test statistic: t=
s12 s22
+
n1 n2
277
Assumptions: The population distributions are (at least approximately) normal and the two samples
are independently selected random samples.
Refer to the Minitab output given in the problem statement.
From the Minitab output the P-value = 0.007. Since the P-value is less than α, Ho is rejected. There
is sufficient evidence to conclude that the mean half-life of vitamin D is longer for those on a normal
diet than for those on a high-fiber diet.
11.26 Let μ1 denote the mean IQ of soccer players who average less than 10 headers per game. Let μ2
denote the mean IQ of soccer players who average 10 or more headers per game
Ho: μ1 − μ 2 = 0 Ha: μ1 − μ 2 > 0
α = 0.05
( x1 − x2 ) − 0
Test statistic: t=
s12 s22
+
n1 n2
n1 = 35, x1 = 112 , s1 = 10, n 2 = 25, x2 = 103, s2 = 8
(112 − 103) − 0 9
t= = = 3.87
(10)2 (8)2 2.327476
+
35 25
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
⎝ n1 n2 ⎠ (2.857143 + 2.56)2
df = 2 2
= = 57.2
1 ⎛ s12 ⎞ 1 ⎛ s22 ⎞ (2.857143)2 (2.56)2
⎜⎜ ⎟⎟ + ⎜ ⎟ +
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎜⎝ n2 ⎟⎠ 34 24
P-value = area under the 57df t curve to the right of 3.87 ≈ 0.00014.
Since the P-value is less than α, Ho is rejected. The sample data supports the conclusion that the
mean IQ of soccer players who frequently head the ball is lower than that of those who do not
frequently head the ball.
11.27 Let μ1 denote the mean self-esteem score for students classified as having short duration loneliness.
Let μ2 denote the mean self-esteem score for students classified as having long duration loneliness.
Ho: μ1 − μ 2 = 0 Ha: μ1 − μ 2 > 0
α = 0.01
278
( x1 − x2 ) − 0
Test statistic: t=
s12 s22
+
n1 n2
n1 = 72, x1 = 76.78 , s1 = 17.8, n 2 = 17, x2 = 64.00, s2 = 15.68
(76.78 − 64.00) − 0 12.78

t= = = 2.9426
2
(17.8) (15.68) 2 4.34316
+
72 17
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
⎝ n1 n2 ⎠ (4.4006 + 14.4625)2
df = 2 2
= = 26.7
1 ⎛ s12 ⎞ 1 ⎛ s22 ⎞ (4.4006)2 (14.4625)2
⎜⎜ ⎟⎟ + ⎜ ⎟ +
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎜⎝ n2 ⎟⎠ 71 16

P-value = area under the 26 df t curve to the right of 2.9426 ≈ 0.0034.
mean self esteem is lower for students classified as having long duration loneliness than for students
classified as having short duration loneliness.
11.28 a. Let μ1 denote the true average oxygen consumption for courting pairs and μ2 the true
average oxygen consumption for non-courting pairs.
Ho: μ1 − μ 2 = 0 Ha: μ1 − μ 2 > 0
α = 0.05
( x1 − x2 ) − 0
Test statistic: t=
sp2 sp2
+
n1 n2
Assumptions: The population distributions are (at least approximately) normal, the two
population standard deviations are equal, and the two samples are independently selected
random samples.
df for pooled t-test = n1 + n2 − 2 = 11 + 15 − 2 = 24
10(0.0066)2 + 14(0.0071)2
sp2 = = 0.000047555
24
(0.099 − 0.072) − 0 0.027
t= = = 9.86
0.000047555 0.000047555 0.002737
+
11 15
Since the P-value is less than α, Ho is rejected. There is sufficient evidence in the sample data to
conclude that the true average oxygen consumption for courting pairs is larger than the true average
oxygen consumption for non-courting pairs.
279
b. Let μ1 denote the true average oxygen consumption for courting pairs and μ2 the true
average oxygen consumption for non-courting pairs.
Ho: μ1 − μ 2 = 0 Ha: μ1 − μ 2 > 0
α = 0.05
( x1 − x2 ) − 0
Test statistic: t=
s12 s22
+
n1 n2
Assumptions: The population distributions are (at least approximately) normal and the two
n1 = 15, x 1 = 0.099 , s1 = 0.0071, n 2 = 11, x 2 = 0.072, s 2 = 0.0066
(0.099 − 0.072) − 0 0.027

t= = = 9.979
2
(0.0071) (0.0066) 2 0.0027057
+
15 11
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
⎝ n1 n2 ⎠ (0.00000336 + 0.00000396)2
df = 2 2
= = 22.57
1 ⎛ s12 ⎞ 1 ⎛ s22 ⎞ (0.00000336)2 (0.00000396)2
⎜ ⎟ + ⎜ ⎟ +
n1 − 1 ⎜⎝ n1 ⎟⎠ n2 − 1 ⎜⎝ n2 ⎟⎠ 14 10
P-value = area under the 22 df t curve to the right of 9.98 ≈ 0.
Since the P-value is less than α, Ho is rejected. There is sufficient evidence in the sample
data to conclude that the true average oxygen consumption for courting pairs is larger than
the true average oxygen consumption for non-courting pairs. So the conclusion is the same
as in part a.
Exercises 11.29 – 11.43
11.29 a. If possible, treat each patient with both drugs with one drug used on one eye and the other
drug used on the other eye. For each patient, determine at random which eye will receive the
new treatment. Then take observations (readings) of eye pressure on each eye. If this
treatment method is not possible, then request the ophthalmologist to pair patients according
to their eye pressure so that the two people in a pair have approximately equal eye pressure.
Then select one patient from each pair to receive the new drug and treat the other patient in
each pair with the standard treatment. Record the reduction in eye pressure. Treat the other
person in that pair with the standard treatment and record the reduction in eye pressure.
These two readings would constitute a pair. Repeat for each of the other pairs to obtain the
paired sample data.
b. Both procedures above would result in paired data.
c. Assign subjects at random to one of the two treatment groups. Measure reduction in eye
pressure for both groups. The resulting observations would constitute independent samples.
This experiment is probably not as informative as a paired experiment with the same number
of subjects to patient to patient variability which can be quite large.
280
11.30 Take n pieces of pipe and cut each into two pieces, resulting in n pairs of pipe. Coat one piece in
each pair with coating 1 and the other piece with coating 2. Then put both pipes from a pair into
service where they are buried at the same depth, orientation, in the same soil type, etc. After the
specified length of time, measure the depth of corrosion penetration for each piece of pipe. The
experiment results in paired data, which "filters out" effects due to the extraneous factors.
11.31 Let d = the difference in the time to exhaustion after chocolate milk and after carbohydrate
replacement drink., i.e. time after chocolate milk – time after carbohydrate replacement drink.
Let μ d denote the mean difference in the time to exhaustion.

Ho: μ d = 0 Ha: μ d > 0
α = 0.05
We must assume that the sample is randomly selected. The sample size is small, but a
boxplot of the differences shows a reasonably symmetrical shape with no outliers. It would
not be unreasonable to assume that the sample data comes from a population that is
approximately normal.
0 5 10 15 20 25 30
difference
xd − 0
The test statistic is: t = with d.f. = 8
sd
n
14.0789 − 0
d = 14.0789 and sd = 9.4745 t= = 4.46
9.4745
9
Since the P-value is much less than α, the null hypothesis should be rejected at 0.05 There
is sufficient evidence to suggest that the mean time to exhaustion is greater after chocolate
milk than after carbohydrate replacement drink.
11.32 Let d = the difference in velocity for each swimmer between swimming through water and swimming
through Guar Syrup, i.e. Water – syrup.
Let μ d denote the mean difference for all swimmers.
Ho: μ d = 0 Ha: μ d ≠ 0 α = 0.01
Assumptions: The sample of swimmers were randomly selected. The sample size is small, and the
boxplot of the differences (below) shows a skewed distribution, but it has no outliers. It would not
be unreasonable to assume that the sample data comes from a population that is approximately
normal.
281
xd − 0
The test -0.050 -0.025 0.000 0.025 0.050 0.075
statistic is: t = with
difference
sd
n
d.f. = 19
d = -0.004 and s d = 0.0347
−0.004 − 0
t= = − 0.52
0.0347
20
P-value = 2(the area under the 19 df t curve to the left of -0.52) ≈ 2(.311) = .622
Since the P-value is greater than α, the null hypothesis cannot be rejected at 0.01 There is
insufficient evidence to suggest that there is any difference in swimming time between swimming in
guar syrup and swimming in water. The authors of this paper were correct.
11.33 Let d = the difference in total body mineral content for each mother between breast feeding (B) and
post weaning. (P), i.e. P - B.
Let μ d denote the mean difference for all women,
Ho: μ d = 25 Ha: μ d > 25

α = 0.05
Assumptions: The sample of mothers are randomly selected. The sample size is small, and the
boxplot of the differences (below) shows a skewed distribution, but it has no outliers. It would not
be unreasonable to assume that the sample data comes from a population that is approximately
normal
282
0 50 100 150 200 250 300 350
difference
xd − 25
sd
n
105.7 − 25
d = 105.7 and s d = 103.845 t= = 2.46
103.845
10
P-value = the area under the 9 df t curve to the right of 2.46 ≈ .017
Since the P-value is less than α, the null hypothesis should be rejected at 0.05 There is sufficient
evidence to suggest that the true average body bone mineral content during post weaning exceeds
that during breast feeding by more than 25 grams.
11.34 a. Let d = the difference in selenium concentration in the milk for each cow given the
selenium supplement both initially and after a 9-day period. i.e. Initially - After
Let μ d denote the mean difference for all cows given the treatment..
Ho: μ d = 0 Ha: μ d < 0
α = 0.05 (no significance level mentioned, so chosen for demonstration)
Assumptions: The sample of cows are randomly selected. The sample size is small, and
the boxplot of the differences (below) shows a skewed distribution, but it has no outliers. It
would not be unreasonable to assume that the sample data comes from a population that is
-160 -150 -140 -130 -120 -110 -100 -90 -80 -70
283
difference
xd − 0
sd
n
d = -104.731 and s d = 24.101, n = 16
−104.731 − 0
t= = − 17.38
24.101
16
P-value = the area under the 15 df t curve to the left of -17.38 ≈ 0.00
Since the P-value is less than α, the null hypothesis should be rejected at 0.05 There is
sufficient evidence to suggest that the true mean selenium concentration is greater after 9
days of selenium supplement.
b. Let d = the difference in selenium concentration in the milk for each cow given a
control both initially and after a 9-day period. i.e. Initially - After
Let μ d denote the mean difference for all cows given the treatment.
Ho: μ d = 0 Ha: μ d ≠ 0
Assumptions: The sample of cows are randomly selected. The sample size is small, and
the boxplot of the differences (below) shows a symmetric distribution. It would not be
unreasonable to assume that the sample data comes from a population that is approximately
normal.
-1 0 1 2 3
difference
xd − 0
sd
n
d = 0.6929 and s d = 1.062, n = 14
0.6929 − 0
t= = 2.44
1.062
14
P-value = 2(the area under the 13 df t curve to the right of 2.44 ≈ 2(.016) = .032
284
sufficient evidence to suggest that the true mean selenium concentration is greater after 9
days with no selenium supplement. Note that this would not have been significant at α =0.01
(or any significance level less than .032) and for these cases, the data would be consistent
with the hypothesis of no change.
c. No the paired t test would not be the correct test to compare the mean difference between
the control group and the treatment group. This would be comparing two different samples of
cows. No cow from the treatment group would be more similar to any cow in the control
group than any other cow in that group, so the two groups are independent. A two sample
test of differences for independent samples would be the correct test. A paired test is used
when there is a link between a data point in one group and a data point in the other (in the
case of parts (a) and (b) in this question, it would be the same cow!
11.35 a. Let μd denote the true average difference in translation between dominant and nondominant
arms for pitchers (dominant – nondominant)
We assume the pitchers were chosen as random and as the same pitcher is used for both
arms, the data is paired. The sample size is small but the boxplot of the differences shows a
fairly symmetrical distribution with one outlier. It is not unreasonable to assume that the
data comes from a population that is approximately normal.
0.0 2.5 5.0 7.5 10.0 12.5 15.0

d =4.066, difference sd = 3.955, n =17.
The 95% confidence interval for μd with df =16 is:
sd ⎛ 3.955 ⎞
d ± (t critical)⇒ 4.066 ± (2.12) ⎜ ⎟
n ⎝ 17 ⎠
⇒ 4.066 ± 2.034 ⇒ (2.032, 6.10).
With 95% confidence, it is estimated that pitchers using their dominant arm have, a greater
average difference in the translation of their shoulder than when they use their nondominant
arm by between 2.032 mm and 6.10 mm.
b. Let μd denote the true average difference in translation between dominant and nondominant
arms for position players (dominant – nondominant)
We assume the players were chosen as random and as the same player is used for both
arms, the data is paired. The sample size is small but the boxplot of the differences shows a
skewed distribution with no outliers. It is not unreasonable to assume that the
data comes from a population that is approximately normal.
285
-3 -2 -1 0 1 2 3 4
difference
d =0.2326, sd = 1.6034, n =19.
The 95% confidence interval for μd with df =18 is:

sd ⎛ 1.6034 ⎞
d ± (t critical)⇒ 0.2326 ± (2.10) ⎜ ⎟
n ⎝ 19 ⎠
⇒ 0.2326 ± 0.7725 ⇒ ( − 0.54, 1.01).
With 95% confidence, it is estimated that position players using their dominant arm have, an
average difference of between 0.54 mm less to 1.01 mm greater, in the translation of their
shoulder, than when they use their nondominant arm.
c. No. This question is asking for the difference in the mean translations for the two types of
players ie: μ pitcher − μ position . This is not the same as the mean of the differences: μ d
11.36 Let d = the difference in radiation measurement on aircraft using two different methods.
i.e. Method 1 – Method 2
Let μ d denote the mean difference for readings.
Ho: μ d = 0 Ha: μ d ≠ 0
Assumptions: It is assumed that the sample of flights are randomly selected. The same
flights were used for both methods so the data is paired. The sample size is small, and the
boxplot of the differences (below) shows a skewed distribution but with no outliers. It would
not be unreasonable to assume that the sample data comes from a population that is
-8 -6 -4 -2 0 2 4
difference
286
xd − 0
sd
n
d = -0.7375 and s d = 3.526, n = 8
−0.7375 − 0
t= = − 0.59
3.526
8
P-value = 2(the area under the 7 df t curve to the left of -0.59 ≈ 2(.284) = .568
Since the P-value is greater than α, the null hypothesis should not be rejected at 0.05 There
is not sufficient evidence to suggest that there is any difference in the true mean radiation
measurement for the two methods,
11.37 a. Let d = the difference in wrist extension while using two different types of computer Mouse.
(Type A – Type B)
Let μ d denote the mean difference for wrist extension between the two types.
Ho: μ d = 0 Ha: μ d > 0
α = 0.05
Assumptions: Each student used both types of mouse so the data is paired. It is stated that
the sample is representative of the population of computer users so it can be seen as a
random sample. The sample size is not large ( n =24) but without the actual data, we must
assume that the population data is approximately normal.
xd − 0
sd
n
d = 8.82 and s d = 10, n = 24
8.82 − 0
t= = 4.32
10
24
sufficient evidence to suggest that the mean wrist extension for mouse A is greater than for
mouse B.
b. The hypotheses and assumptions will be the same as in Part (a).
xd − 0
sd
n
d = 8.82 and s d = 25, n = 24

287
8.82 − 0
t= = 1.73
25
24
Although the P-value is just greater than α , the null hypothesis should be not be rejected.
However, there is certainly not “convincing evidence” that the mean wrist extension for
mouse A is greater than for mouse B.
c. Changing the standard deviation of the differences made the variation in the distribution of
the differences greater. So, for the same value of the sample mean difference, the area in
the tail was larger and the P-value was greater.
11.38 a. Let d = the difference in the percent of exams earning college credit at each central coach
high school i.e. % in 1997 - % in 2002.
Let μd denote the mean difference in the percent of exams earning college credit at each
central coach high school i.e. % in 1997 - % in 2002
Ho: μ d = 0 Ha: μ d > 0
α = 0.05 (not stated in the question, but used for demonstration purposes)
Assumptions: The sample is random and independent. The sample size is small, but a
boxplot of the differences shows symmetry, suggesting that distribution of the population
differences are approximately normal.
x −0
The test statistic is: t = d with d.f. = 6
sd
n
d = -5.4 and sd = 12.05
−5.4 − 0
t= = − 1.15
12.05
7
P-value = the area under the 6 df t curve to the right of -1.15) ≈ 0.853.
Thus, the null hypothesis cannot be rejected at 0.05 There is not sufficient evidence to
suggest that the mean difference in the percent of exams earning college credit at each
central coach high school has declined between 1997 and 2002.
b. No. These 7 schools are representative of high schools located on the central coast of
California, but they are not representative of all California high schools.
c. After computing the differences, there is clearly an outlier (school #5). With such a small
sample size, this violates one of the assumptions of the t-test. It would be doubtful that the
sample of differences would come from an approximately normal population and therefore a
t-test would not be appropriate.
288
11.39 Let d = the difference in the cost-to-charges ratio for both inpatient and outpatient care.
Let μ d denote the mean difference in the ratio for 6 hospitals in Oregon in 2002
i.e. Inpatient - Outpatient
Ho: μ d = 0 Ha: μ d > 0
α = 0.05
Assumptions: The sample is random and independent. The sample size is small, but a boxplot of the
differences shows reasonable symmetry, suggesting that distribution of the population differences are
xd − 0
sd
n
d = 18.833 and s d = 5.6716
18.833 − 0
t= = 8.13
5.6716
6
P-value = the area under the 5 df t curve to the right of 8.13) ≈ 0.
Thus, the null hypothesis is rejected at 0.05 There is sufficient evidence to suggest that the mean
difference cost-to-charge ratio for Oregon hospitals is lower for outpatient care than for inpatient care.
11.40 Let μ d denote the mean difference in verbal ability in children born prematurely (aged 8 – aged 3)
Ho: μ d = 0 Ha: μd > at or above the proficient level in mathematics 0
α = 0.05 (not stated, but used for demonstration)
From MINITAB: t = 3.17, P-value = 0.001
Thus, the null hypothesis is rejected. There is sufficient evidence to suggest that the mean verbal
ability in children born prematurely increases between the ages of 3 and 8.
11.41 It is not necessary to use an inference procedure since complete information on all 50 states is
available. Inference is necessary only when a sample is selected from some larger population.
11.42 a. The data are paired because the response for the number of science courses each girl in the
sample intended to take is logically matched with the same girl’s response for the number of
science courses she thought boys should take.
b. Let μd denote the true average difference in the intended number of courses for girls and
boys (girls – boys).
The 95% confidence interval for μd is

s ⎛ 1.51 ⎞
d ± (t critical) d ⇒ − 0.83 ± (1.971) ⎜ ⎟
n ⎝ 223 ⎠
⇒ − 0.83 ± 0.1988 ⇒ ( − 1.029, − 0.631).
With 95% confidence, it is estimated that the mean difference in the number of science
courses girls intend to take and what they think boys should take is −1.029 and −0.631.
289
11.43 a. Let μ d denote the mean change in blood lactate level for male racquetball players
(After − Before).
The differences are: 5, 17, 23, 22, 17, 4, 18, 3.
From these, x d = 13.625 and s d = 8.2797
The 95% confidence interval for μd is
sd ⎛ 8.2797 ⎞
d ± (t critical) ⇒ 13.625 ± (2.365) ⎜ ⎟
n ⎝ 8 ⎠
⇒ 13.625 ± 6.923 ⇒ (6.702, 20.548).
With 95% confidence, it is estimated that the mean change in blood lactate level for male
racquetball players is between 6.702 and 20.548.
b. Let μ d denote the mean change in blood lactate level for female racquetball players
(After − Before).
The differences are: 10, 10, 6, 3, 0, 20, 7.
From these, x d = 8.0 and s d = 6.4031
The 95% confidence interval for μ d is

sd ⎛ 6.4031 ⎞
d ± (t critical)⇒ 8.0 ± (2.45) ⎜ ⎟
n ⎝ 7 ⎠
⇒ 8.0 ± 5.929 ⇒ (2.071, 13.929).
With 95% confidence, it is estimated that the mean change in blood lactate level for female
racquetball players is between 2.071 and 13.929.
c. Since the two intervals overlap (have values in common), this suggests that it is possible for
the mean change for males and the mean change for females to have the same value. (It is
appropriate to examine a confidence interval for the difference between the mean changes
for men and women to answer this question.)
Exercises 11.44 – 11.60
11.44 Let π1 denote the proportion of students who registered by phone that were satisfied with the
registration process and π2 denote the corresponding proportion for those who registered on-line.
Ho: π 1 − π 2 = 0 Ha: π 1 − π 2 < 0
α = 0.05
p1 − p 2
z=
p c(1 − p c ) p c(1 − p c )
+
n1 n2
57 50
p1 = = 0.7125 p2 = = 0.8333
80 60
290
n1p1 + n2 p2 57 + 50
pc = = = 0.7643
n1 + n2 80 + 60
(0.7125 − 0.8333) −0.1208

z= = = −1.666
0.7643(1 − 0.7643) 0.7643(1 − 0.7643) 0.0725
+
80 60
P-value = Area under the z curve to the left of –1.666 = 0.0479.
Since the P-value is less than α, Ho is rejected. The data supports the claim that the proportion of
satisfied students is higher for those who registered on-line than for those who registered over the
phone.
11.45 Let π1 denote the proportion of passengers that flew on airplanes that did not recirculated air that
reported post-flight respiratory symptoms, and π2 denote the proportion of passengers that flew on
airplanes that did recirculated air that reported post-flight respiratory symptoms
Ho: π 1 − π 2 = 0 Ha: π 1 − π 2 ≠ 0
α = 0.05
p1 − p 2
z=
p c(1 − p c ) p c(1 − p c )
+
n1 n2
108 111
p1 = = 0.2089 p2 = = 0.1904
517 583
n1p1 + n2 p2 108 + 111

pc = = = 0.1991
n1 + n2 517 + 583
(0.2089 − 0.1904) 0.0185

z= = =0.7676
0.1991(1 − 0.1991) 0.1991(1 − 0.1991) 0.0241
+
517 583
P-value = 2(Area under the z curve to the right of 0.77) = 2(0.2206) = 0.4412
Since the P-value is greater than α, Ho is not rejected. There is not enough evidence to suggest that
the proportion of passengers that reported post-flight respiratory symptoms differs for planes that do
and do not recirculate air.
11.46 Let π 1 denote the proportion of patients receiving the stocking treatment who improve and
π 2 denote the proportion of patients receiving the standard treatment that improve.
H0 : π 1 − π 2 = 0 Ha π1 − π 2 > 0
α = 0.05
p1 − p 2
z=
p c(1 − p c ) p c(1 − p c )
+
n1 n2
p1 = .38 n = 57 p2 = .27, n = 50
291
The subjects were assigned randomly to the two treatments. Both samples are large:
n1 p1 = 57(.38) = 21.66 ≥ 10, n1 (1 − p1 ) = 57(.62) = 35.34 ≥ 10
n2 p2 = 50(.27) = 13.5 ≥ 10, n2 (1 − p2 ) = 50(.73) = 36.5 ≥ 10
n1p1 + n2 p2 21.66 + 13.5

pc = = = 0.3286
n1 + n2 57 + 50
(0.38 − 0.27) 0.11

z= = = 1.21
0.3286(0.6714) 0.3286(0.6714) 0.0910
+
57 50
Since the P-value is greater than α, H 0 cannot be rejected. The data does not support the claim the
there is convincing evidence that the proportion of patients who improve is higher for the experimental
treatment than for the standard treatment.
11.47 Let π 1 denote the proportion of Americans aged 12 or older who own an MP3 player in 2006 and
π 2 denote the proportion of Americans aged 12 or older who own an MP3 player in 2005.
n1 = 1112, p1 = 0.20, n2 = 1112, p 2 = 0.15
The samples are independently selected random samples and both sample sizes are large:
n1 p1 = 1112(.20) = 222.4 ≥ 10, n1 (1 − p1 ) = 1112(.80) = 889.6 ≥ 10
n2 p2 = 1112(.15) = 166.8 ≥ 10, n2 (1 − p2 ) = 1112(.85) = 945.2 ≥ 10
p1(1 − p1 ) p2 (1 − p2 )
The 95% confidence interval for π 1 − π 2 is: ( p1 − p2 ) ± zcrit +
n1 n2
0.20(1 − 0.20) 0.15(1 − 0.15)
(0.20 − 0.15) ± 1.96 + ⇒ 0.05 ± 1.96(0.0161)
1112 1112
⇒ 0.05 ± 0.0316 ⇒ (0.0184, 0.0816).
We estimate that the proportion of all Americans the age of 12 or more that owns an MP3 player is
between 1.84% and 8.2% higher than the proportion in 2005. We used a method to construct this
estimate that captures the true difference in the proportions 95% of the time in repeated sampling.
Zero is not included in the interval indicating that the proportion of Americans age 12 and older who
own an MP3 player was significantly higher in 2006 than in 2005.
11.48 Let π 1 denote the proportion of males in the previous study who owned an MP3 player in 2006 and
π 2 denote the proportion of females in the previous study who owned an MP3 player in 2006.
H0 : π 1 − π 2 = 0 Ha π1 − π 2 > 0
α = 0.01
p1 − p 2
z=
p c(1 − p c ) p c(1 − p c )
+
n1 n2
p1 = .24 n = 556 p2 = .16, n = 556
292
The two samples were independently chosen samples. Both samples are large:
n1 p1 = 556(.24) = 133.44 ≥ 10, n1 (1 − p1 ) = 556(.76) = 422.56 ≥ 10
n2 p2 = 556(.16) = 88.96 ≥ 10, n2 (1 − p2 ) = 556(.84) = 467.04 ≥ 10
n1p1 + n2 p2 133.44 + 88.96

pc = = = 0.2
n1 + n2 556 + 556
(0.24 − 0.16) .08

z= = = 3.33
0.2(0.8) 0.2(0.8) 0.024
+
556 556
Since the P-value is less than α, H 0 should be rejected. There is convincing evidence that the
proportion of female Americans aged 12 and over that owned an MP3 player in 2006 is smaller than
the corresponding proportion for males.
11.49 Let π 1 denote the proportion of subjects who experienced gastrointestinal symptoms after eating
olestra chips and π 2 denote the proportion of subjects who experienced gastrointestinal symptoms
after eating regular chips.
H0 : π 1 − π 2 = 0 H a π1 − π 2 ≠ 0
α = 0.05
p1 − p 2
z=
p c(1 − p c ) p c(1 − p c )
+
n1 n2
90 93
p1 = =0.160 p2 = =0.176
563 529
The treatments were assigned at random.. Both samples are large:
n1 p1 = 563(.16) = 90 ≥ 10, n1 (1 − p1 ) = 563(.84) = 473 ≥ 10
n2 p2 = 529(.176) = 93 ≥ 10, n2 (1 − p2 ) = 529(.824) = 436 ≥ 10
n1p1 + n2 p2 90 + 93
pc = = = 0.168
n1 + n2 563 + 529
(0.16 − 0.176) −0.016

z= = = −0.71
0.168(0.832) 0.168(0.832) .0226
+
563 529
P-value = 2(area under the z curve to the left of -0.71 = 2(0.2389) = .4778.
Since the P-value is greater than α, H 0 should not be rejected. There is no evidence that the
proportion of individuals who experience gastrointestinal symptoms after eating olestra chips differs
from the proportion who experience symptoms after consuming regular ones.
293
11.50 Let π 1 denote the proportion of students in grades 6-12 who regard science and math skills crucial
and π 2 denote the proportion of parents who regard science and math skills crucial.
H0 : π 1 − π 2 = 0 Ha π1 − π 2 ≠ 0
α = 0.05
p1 − p 2
z=
p c(1 − p c ) p c(1 − p c )
+
n1 n2
p1 = 0.50 n = 1342 p2 = 0.62, n = 1379
The two samples were independently chosen samples. Both samples are large:
n1 p1 = 1342(.5) = 671 ≥ 10, n1 (1 − p1 ) = 1342(.5) = 671 ≥ 10
n2 p2 = 1379(.62) = 854.98 ≥ 10, n2 (1 − p2 ) = 1379(.38) = 524.02 ≥ 10
n1p1 + n2 p2 671 + 854.98

pc = = = 0.561
n1 + n2 1342 + 1379
(0.5 − 0.62) −.12

z= = = −6.32
0.561(0.439) 0.561(0.439) 0.019
+
1342 1379
P-value = 2(area under the z curve to the left of -6.32) ≈ 0.0
proportion of parents who regard science and mathematics as crucial is different than the
corresponding proportion for students in grades 6-12.
11.51 a. Let π 1 denote the proportion of cardiologists who do not know that carbohydrate was the
diet component most likely to raise triglycerides and π 2 be the corresponding proportion for
internists.
26 222
p1 = = 0.217, p2 = = 0.530
120 419
The samples are independently selected samples and both sample sizes are large:
n1 p1 = 120(.217) = 26 ≥ 10, n1 (1 − p1 ) = 120(.783) = 94 ≥ 10
n2 p2 = 419(.53) = 222 ≥ 10, n2 (1 − p2 ) = 419(.47) = 197 ≥ 10
Randomness is discussed in part (b)
p1(1 − p1 ) p2 (1 − p2 )
n1 n2
0.217(1 − 0.217) 0.53(1 − 0.53)

(0.217 − 0.530) ± 1.96 + ⇒ -0.313 ± 1.96(0.0448)
120 419
⇒ -0.313 ± 0.088 ⇒ (-0.401, -0.225).
294
We estimate that that the proportion of all cardiologists who do not know that carbohydrate
was the diet component most likely to raise triglycerides is between 22.5% and 40% lower
than the corresponding internists. We used a method to construct this estimate that captures
the true difference in the proportions 95% of the time in repeated sampling.
. This is a volunteer sample – the physicians were asked to reply to the questionnaire and 84%
of them did not; a very high non-response rate. It may be that of the cardiologists, only those
who knew the answers to the questions, sent back the questionnaire and the internists sent
them back whether they knew the answers or not.
11.52 a. As the vaccinations are delivered by two different methods, it would be obvious to both the
patients and the administrators which patient had which vaccine and may influence their
response.
b. Let π 1 denote the proportion of children who get sick after receiving the vaccine by
nasal spray and π 2 be the corresponding proportion for children receiving the vaccine by
injection.
p1 = 0.039 n = 4000 p2 = 0.086, n = 4000
We assume the samples are independently selected random samples and both sample sizes
are large:
n1 p1 = 4000(.039) = 156 ≥ 10, n1 (1 − p1 ) = 4000(.961) = 3844 ≥ 10
n2 p2 = 4000(.086) = 344 ≥ 10, n2 (1 − p2 ) = 4000(.914) = 3656 ≥ 10
p1(1 − p1 ) p2 (1 − p2 )
n1 n2
0.039(1 − 0.039) 0.086(1 − 0.086)

(0.039 − 0.086) ± 2.58 + ⇒ -0.047 ± 2.58(0.0054)
4000 4000
⇒ -0.047 ± 0.014 ⇒ (-0.061, -0.033).
We estimate that the proportion of children who get sick after receiving the vaccine by nasal
spray is between 3.3% and 6.1% less than the children who receive the vaccine by injection.
We used a method to construct this estimate that captures the true difference in the
proportions 99% of the time in repeated sampling.
11.53 Let π 1 denote the proportion of all college graduates than get a sunburn and π 2 be the proportion of
all those without a high school degree that than get a sunburn.
H0 : π 1 − π 2 = 0 Ha π1 − π 2 > 0
α = 0.05
p1 − p 2
z=
p c(1 − p c ) p c(1 − p c )
+
n1 n2
p1 = .43 n = 200 p2 = .25, n = 200
295
We assume the samples are independently selected random samples and both sample sizes are
large:
n1 p1 = 200(.43) = 86 ≥ 10, n1 (1 − p1 ) = 200(.57) = 114 ≥ 10
n2 p2 = 200(.25) = 50 ≥ 10, n2 (1 − p2 ) = 200(.75) = 150 ≥ 10
n1p1 + n2 p2 86 + 50
pc = = = 0.68
n1 + n2 200 + 200
(0.43 − 0.25) .18

z= = = 3.86
0.68(0.32) 0.68(0.32) .0466
+
200 200
proportion who experience a sunburn is higher for college graduates than it is for those without a high
school degree.
11.54 a. Let π 1 denote the proportion of all Americans reporting major depression and π 2 denote
the proportion of all Canadians reporting major depression
H0 : π 1 − π 2 = 0 Ha π1 − π 2 > 0
α = 0.05
p1 − p 2
z=
p c(1 − p c ) p c(1 − p c )
+
n1 n2
p1 = .09 (assume) n = 5183 p2 = .082 (known), n = 3505
are large:
n1 p1 = 5183(.09) = 466.47 ≥ 10, n1 (1 − p1 ) = 5183(.91) = 4716.53 ≥ 10
n2 p2 = 3505(.082) = 287.41 ≥ 10, n2 (1 − p2 ) = 3505(.918) = 3217.59 ≥ 10
n1p1 + n2 p2 466.47 + 287.41

pc = = = 0.087
n1 + n2 5183 + 3505
(0.09 − 0.082) .007

z= = = 1.17
0.087(0.913) 0.087(0.913) .006
+
5183 3505
Since the P-value is greater than α, H 0 should not be rejected. Yes, the proportion of
Americans reporting major depression could have been as large as .09 as this difference is
not statistically significant.
296
b. The hypotheses and significance level are the same as in part (a).
p1 = .10 (assume) n = 5183 p2 = .083 (known), n = 3505

are large:
n1 p1 = 5183(.10) = 518.3 ≥ 10, n1 (1 − p1 ) = 5183(.9) = 4664.7 ≥ 10
n2 p2 = 3505(.083) = 290.915 ≥ 10, n2 (1 − p2 ) = 3505(.917) = 3214.085 ≥ 10
n1p1 + n2 p2 518.3 + 290.915

pc = = = 0.093
n1 + n2 5183 + 3505
(0.10 − 0.083) .017

z= = = 2.83
0.093(0.907) 0.093(0.907) .006
+
5183 3505
Since the P-value is less than α, H 0 can be rejected. No, the proportion of Americans
reporting major depression could not have been as large as .10 as this difference is
statistically significant.
11.55 Let π 1 denote the proportion of exposed dogs that develop lymphoma and π 2 be the
corresponding proportion for unexposed dogs.
p1 = 0.572 n = 827 p2 = 0.146, n = 130
are large:
n1 p1 = 827(.572) = 473.044 ≥ 10, n1 (1 − p1 ) = 827(.428) = 353.956 ≥ 10
n2 p2 = 130(.146) = 18.98 ≥ 10, n2 (1 − p2 ) = 130(.954) = 124.02 ≥ 10
p1(1 − p1 ) p2 (1 − p2 )
n1 n2
0.572(1 − 0.428) 0.146(1 − 0.146)

(0.572 − 0.146) ± 1.96 + ⇒ .426 ± 1.96(0.0354)
827 130
⇒ .426 ± .069 ⇒ (.357, .495).
We believe that the proportion of exposed dogs that develop lymphoma exceeds that for
unexposed dogs by somewhere between .357 and .495. We used a method to construct this
estimate that captures the true difference in the proportions 95% of the time in repeated
sampling.
297
11.56 Let π1 denote the proportion of resumes with white sounding names receiving positive responses, and
π2 denote the proportion of resumes with black sounding names receiving positive responses
Ho: π 1 − π 2 = 0 Ha: π 1 − π 2 > 0
p1 − p 2
z=
p c(1 − p c ) p c(1 − p c )
+
n1 n2
250 167
p1 = = 0.1 p2 = = 0.0668
2500 2500
n1p1 + n2 p2 250 + 167

pc = = = 0.0844
n1 + n2 5000
(0.1 − 0.0668) 0.0332

z= = =4.25
0.0834(1 − 0.0834) 0.0834(1 − 0.0834) 0.0078
+
2500 2500
P-value = Area under the z curve to the right of 4.25 ≈ 0
Since the P-value is less than α, Ho is rejected. There is enough evidence to suggest that the
proportion receiving positive responses is higher for those resumes with white-sounding first names.
11.57 a. Let π1 denote the proportion of Austrian avid mountain bikers with low sperm counts, and π2
denote the proportion of Austrian non-bikers with low sperm counts.
Ho: π 1 − π 2 = 0 Ha: π 1 − π 2 > 0
p1 − p 2
z=
p c(1 − p c ) p c(1 − p c )
+
n1 n2
p1 = 0.9 p2 = 0.26
n1p1 + n2 p2 90 + 26
pc = = = 0.58
n1 + n2 200
(0.9 − 0.26) 0.64

z= = =9.17
0.58(1 − 0.58) 0.58(1 − 0.58) 0.0698
+
100 100
P-value = Area under the z curve to the right of 9.17 ≈ 0
Since the P-value is less than α, Ho is rejected. There is enough evidence to suggest that the
proportion of Austrian avid mountain bikers with low sperm count is higher than the proportion
of Austrian non-bikers.
298
b. These were not a group of men who where randomly put into one of two treatment groups. It
was an observational study and no cause and effect conclusion can be made from such a
study.
11.58 a. Let π1 denote the true proportion of all high risk patients who receive insulin and develop
diabetes and π2 denote the true proportion of all high risk patients who receive do not
receive insulin and develop diabetes
25 24
n1 = 169, x1 = 25, p1 = = 0.1479, n2 = 170, x 2 = 24, p2 = = 0.1412
169 170
The 90% confidence interval for π 1 − π 2 is
0.1479(0.8521) 0.1412(0.8588)
(0.1479 − 0.1412) ± 1.645 + ⇒ 0.0067 ± 1.645(0.0382)
169 170
⇒ 0.0067 ± 0.0628 ⇒ ( − 0.0561, 0.0695).
b. With 90% confidence, it is estimated that the true proportion of patients developing diabetes
may be as much as 0.0695 more in the insulin group than in the control group; but it also may
be as much as 0.0561 less in the insulin group as in the control group.
c. Because 0 is in the interval, it is possible that there is no difference in the proportion of the
patients developing diabetes in the two groups. The proposed treatment doesn’t appear very
effective.
11.59 The researchers looked at the hypotheses: H0: π 1 − π 2 = 0 vs. Ha π 1 − π 2 ≠ 0 where π1 is the
proportion of women who survived after 20 years having had a mastectomy and π2 is the proportion of
women who survived after 20 years having had a lumpectomy and radiation. If they reported no
significant difference, they would have failed to reject the null hypothesis.
11.60 The decision was based on all an analysis of all the soldiers that went served in the Gulf War (the
population). Because a census was performed, no inference procedure was necessary.
Exercises 11.61 – 11.90
11.61 a. Let μ1 be the mean elongation (mm) for the square knot and μ2 the mean elongation for
the Duncan loop when using Maxon thread.
Ho: μ1 − μ 2 = 0 Ha: μ1 − μ 2 ≠ 0
( x1 − x2 ) − 0
Test statistic: t=
2 2
s1 s 2
+
n1 n 2
n1 = 10, x 1 = 10.0, s1 = 0.1, n 2 = 15, x 2 = 11.0, s2 = 0.3
(10.0 − 11.0) − 0 −1.0

t= = = 11.9523
2
(0.1) (0.3) 2 0.083666
+
10 15
299
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
⎝ n1 n2 ⎠ (0.001 + 0.006)2
df = 2 2
= = 18.27
1 ⎛ s12 ⎞ 1 ⎛ s22 ⎞ (0.001)2 (0.006)2
⎜⎜ ⎟⎟ + ⎜ ⎟ +
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎜⎝ n2 ⎟⎠ 9 14
P-value = 2 times the area under the 18 df t curve to the right of 11.9 ≈ 0.
Since the P-value is less than α, the null hypothesis is rejected. The data supports the
conclusion that the true mean elongation for the square knot and the Duncan loop differ when
using Maxon thread.
b. Let μ1 be the mean elongation (mm) for the square knot and μ2 the mean elongation for
the for the Duncan loop for the Ticon thread.
Ho: μ1 − μ 2 = 0 Ha: μ1 − μ 2 ≠ 0
( x1 − x2 ) − 0
Test statistic: t=
2 2
s1 s 2
+
n1 n 2
n1 = 10, x 1 = 2.5, s1 = 0.06, n 2 = 11, x 2 = 10.9, s2 = 0.4
(2.5 − 10.9) − 0 −8.4

t= = = −68.796
2
(0.06) (0.4) 2 0.1221
+
10 11
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
⎝ n1 n2 ⎠ (0.00036 + 0.014545)2
df = 2 2
= = 10.494
1 ⎛ s12 ⎞ 1 ⎛ s22 ⎞ (0.00036)2 (0.014545)2
⎜⎜ ⎟⎟ + ⎜ ⎟ +
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎜⎝ n2 ⎟⎠ 9 10
P-value = 2 times the area under the 10 df t curve to the left of –68.8 ≈ 0.
Since the P-value is less than α, the null hypothesis can be rejected. The data supports the
conclusion that the true mean elongation for the square knot and the Duncan loop differ when
using Ticron thread.
c. Let μ1 be the mean elongation (mm) for the Maxon thread and μ2 the mean elongation for the
Ticron thread when using the Duncan loop.
Ho: μ1 − μ 2 = 0 Ha: μ1 − μ 2 ≠ 0
300
( x − x2 ) − 0
Test statistic: t= 1
2 2
s1 s 2
+
n1 n 2
n1 = 15, x 1 = 11.0, s1 = 0.3, n 2 = 11, x 2 = 10.9, s2 = 0.4

(11.0 − 10.9) − 0 0.1
t= = = 0.6977
2
(0.3) (0.4) 2 0.143337
+
15 11
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
⎝ n1 n2 ⎠ (0.006 + 0.014545)2
df = 2 2
= = 17.790
1 ⎛ s12 ⎞ 1 ⎛ s22 ⎞ (0.006)2 (0.014545)2
⎜⎜ ⎟⎟ + ⎜ ⎟ +
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎜⎝ n2 ⎟⎠ 14 10
Since the P-value is greater than α of 0.01 (or even 0.05), the null hypothesis cannot be
rejected. The data do not indicate that there is a difference between the mean
elongations for the Maxon thread and the Ticron thread when using the Duncan loop.
11.62 Small prey

Let μ1 be the mean amount (mg) of venom injected by the inexperienced snakes and μ2 the
mean amount of venom injected by the experienced snakes when the prey is a small prey.
Ho: μ1 − μ 2 = 0 Ha: μ1 − μ 2 ≠ 0
( x − x2 ) − 0
2 2
s1 s 2
+
n1 n 2
n1 = 7, x 1 = 3.1, s1 = 1.0, n 2 = 7, x 2 = 2.6, s2 = 0.3
(3.1 − 2.6) − 0 0.5

t= = = 1.2670
2
(1.0) (0.3) 2 0.3946
+
7 7
301
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
⎝ n1 n2 ⎠ (0.1428 + 0.0128)2
df = 2 2
= = 7.0713
1 ⎛ s12 ⎞ 1 ⎛ s22 ⎞ (0.1428)2 (0.0128)2
⎜⎜ ⎟⎟ + ⎜ ⎟ +
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎜⎝ n2 ⎟⎠ 6 6
Since the P-value is greater than α, the null hypothesis cannot be rejected. At level of
significance of 0.05 (or even 0.10), the data do not indicate that there is a difference in the
amount of venom injected between inexperienced snakes and experienced snakes when the
prey is a small prey.
Medium prey:
mean amount of venom injected by the experienced snakes for medium prey.
Ho: μ1 − μ 2 = 0 Ha: μ1 − μ 2 ≠ 0
( x − x2 ) − 0
2 2
s1 s 2
+
n1 n 2
n1 = 7, x 1 = 3.4, s1 = 0.4, n 2 = 7, x 2 = 2.9, s2 = 0.6
(3.4 − 2.9) − 0 0.5

t= = = 1.8344
2
(0.4) (0.6) 2 0.2725
+
7 7
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
⎝ n1 n2 ⎠ (0.0228 + 0.0514)2
df = 2 2
= = 10.45
1 ⎛ s12 ⎞ 1 ⎛ s22 ⎞ (0.0228)2 (0.0514)2
⎜⎜ ⎟⎟ + ⎜ ⎟ +
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎜⎝ n2 ⎟⎠ 6 6
Since the P-value is greater than α, the null hypothesis cannot be rejected. At level of
significance 0.05, the data do not indicate that there is a difference in the amount of venom
injected between inexperienced snakes and experienced snakes for medium prey.
302
Large prey:
mean amount of venom injected by the experienced snakes for large prey.
Ho: μ1 − μ 2 = 0
Ha: μ1 − μ 2 ≠ 0
( x1 − x2 ) − 0
Test statistic: t=
2 2
s1 s 2
+
n1 n 2
n1 = 7, x 1 = 1.8, s1 = 0.3, n 2 = 7, x 2 = 4.7, s2 = 0.3
(1.8 − 4.7) − 0 −2.9

t= = = −18.0846
2
(0.3) (0.3) 2 0.1603
+
7 7
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
⎝ n1 n2 ⎠ (0.0128 + 0.0128)2
df = 2 2
= = 12.0 . So df = 12
1 ⎛ s12 ⎞ 1 ⎛ s22 ⎞ (0.0128)2 (0.0128)2
⎜ ⎟ + ⎜ ⎟ +
n1 − 1 ⎜⎝ n1 ⎟⎠ n2 − 1 ⎜⎝ n2 ⎟⎠ 6 6
Since the P-value is smaller than α, the null hypothesis is rejected. The data provide
strong evidence that there is a difference in the amount of venom injected between
inexperienced snakes and experienced snakes for large prey.
11.63 Let μ1 be the mean relative area of orange for Yarra guppies and μ2 the mean relative area of
orange for Paria guppies.
Ho: μ1 − μ 2 = 0 Ha: μ1 − μ 2 ≠ 0
( x1 − x2 ) − 0
Test statistic: t=
2 2
s1 s 2
+
n1 n 2
n1 = 30, x 1 = 0.106, s1 = 0.055, n 2 = 30, x 2 = 0.178, s2 = 0.058
303
(0.106 − 0.178) − 0 −0.0720
t= = = −4.9337
2
(0.055) (0.058) 2 0.0145
+
30 30
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
⎝ n1 n2 ⎠ (0.000101 + 0.000112)2
df = 2 2
= = 57.8
1 ⎛ s12 ⎞ 1 ⎛ s22 ⎞ (0.000101)2 (0.000112)2
⎜⎜ ⎟⎟ + ⎜ ⎟ +
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎜⎝ n2 ⎟⎠ 29 29
Since the P-value is smaller than α, the null hypothesis is rejected. The data provide strong evidence
that there is a difference in the amount of relative area of orange between Yarra guppies and Paria
guppies with Yarra guppies having a lower mean relative area of orange.
11.64 a. Let μ1 denote the true mean level of testosterone for male trial lawyers and μ2 the true mean
level of testosterone for male nontrial lawyers. We wish to test the null hypothesis
Ho: μ1 − μ 2 = 0 against the alternative hypothesis Ha: μ1 − μ 2 ≠ 0. The t statistic for this test
is reported to be 3.75 and the degrees of freedom are 64. The P-value is twice the area
under the 64 df t curve to the right of 3.75. This is equal to 0.0004 (the report states that the
P-value is < 0.001 but doesn’t report the actual P-value). Hence the data do provide strong
evidence to conclude that the mean testosterone levels for male trial lawyers and nontrial
lawyers are different.
b. Let μ1 denote the true mean level of testosterone for female trial lawyers and μ2 the true
mean level of testosterone for female nontrial lawyers. We wish to test the null hypothesis
Ho: μ1 − μ 2 = 0 against the alternative hypothesis Ha: μ1 − μ 2 ≠ 0. The t-statistic for this test
is reported to be 2.26 and the degrees of freedom are 29. The P-value is twice the area
under the 29 df t curve to the right of 2.26. This is equal to 0.0316 (the report states that the
P-value is < 0.05 but doesn’t report the actual P-value). Hence the data do provide sufficient
evidence to conclude that the mean testosterone levels for female trial lawyers and female
nontrial lawyers are different.
c. There is not enough information to carry out a test to determine whether there is a significant
difference in the mean testosterone levels of male and female trial lawyers. To carry out
such a test we need the sample means and sample standard deviations for the 35 male trial
lawyers and the 13 female trial lawyers.
11.65 a. Let μ1 denote the true mean “campus involvement” score for returning students and μ2 be the
corresponding score for the nonreturning students. Then μ1 − μ2 denotes the difference
between the means of campus involvement scores for returning and nonreturning students.
n1 = 48, x1 = 3.21, s1 = 1.01, n2 = 42, x2 = 3.31, s2 = 1.03
s12 (1.01)2 s22 (1.03)2

V1 = = = 0.02125 V2 = = = 0.02526
n1 48 n2 42
(V1 + V2 )2 (0.02125 + 0.02526)2

df = = = 85.94
V12 V2 (0.02125)2 (0.02526)2
+ 2 +
n1 − 1 n2 − 1 47 41
304
Use df = 85. The t critical value is 1.988.
The 95% confidence interval for μ1 − μ2 based on this sample is
(3.21 − 3.31) ± 1.988 0.02125 + 0.02526 ⇒ -0.1 ± 1.988(0.2157) ⇒ ( − 0.529,0.329) .
Observe that the interval includes 0, and so 0 is one of the plausible values of μ1 − μ2. That
is, it is plausible that there is no difference between the mean campus involvement scores for
returning and nonreturning students.
b. Let μ1 be the true mean “personal contact” score for returning students and μ2 be the
corresponding score for the nonreturning students.
Ho: μ1 − μ 2 = 0 Ha: μ1 − μ 2 > 0 (i.e., mean score for nonreturning students is lower)
α = 0.01
( x1 − x2 ) − 0
Test statistic: t=
2 2
s1 s 2
+
n1 n 2
n1 = 48, x 1 = 3.22, s1 = 0.93, n1 = 42, x 2 = 2.41, s2 = 1.03
(3.22 − 2.41) − 0 0.81

t= = = 3.894
2
(0.93) (1.03) 2 0.208
+
48 42
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
⎝ n1 n2 ⎠ (0.01802 + 0.02526)2
df = 2 2
= = 83.36
1 ⎛ s12 ⎞ 1 ⎛ s22 ⎞ (0.01802)2 (0.02526)2
⎜⎜ ⎟⎟ + ⎜ ⎟ +
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎜⎝ n2 ⎟⎠ 47 41
Since the P-value is less than α, the null hypothesis of no difference is rejected. The data
provide strong evidence to conclude that the mean “personal contact” score for non-returning
students is lower than the corresponding score for the returning students.
305
11.66 a. Let μ denote the mean salary (in Canadian dollars) for the population of female MBA
graduates of this Canadian business school.
Ho: μ = 100,000 Ha: μ > 100,000
A value for α was not specified in the problem. We will compute the P-value.
x − 100,000
s
n
105,156 − 100,000 5156.0

t= = = 0.7988
98,525 6454.587
233
P-value = area under the 232 d.f. t curve to the right of 0.7988 ≈ 0.2126
For significance levels greater than 0.2126 we can conclude that the mean salary of female
MBA graduates from this business school is above 100,000 dollars.
b. Let μ1 be the true mean salary for female MBA graduates from this business school and
μ2 be the mean for male MBA graduates.
Ho: μ1 − μ 2 = 0 Ha: μ1 − μ 2 < 0
α = 0.01 (a value for α is not specified in this problem. We will use α = 0.01 for
illustration.)
( x1 − x2 ) − 0
Test statistic: t=
2 2
s1 s 2
+
n1 n 2
n1 = 233, x 1 = 105,156, s1 = 98,525, n 2 = 258, x 2 = 133,442, s2 = 131,090
(105,156 − 133,442) − 0 −28,286

t= = = −2.718
(98,525)2 (131,090)2 10405.22
+
233 258
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
⎝ n1 n2 ⎠ (41,661,697.96 + 66,606,930.62)2
df = 2 2
= = 473.7
1 ⎛ s12 ⎞ 1 ⎛ s22 ⎞ (41,661,697.96)2 (66,606,930.62)2
⎜⎜ ⎟⎟ + ⎜ ⎟ +
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎜⎝ n2 ⎟⎠ 232 257
306
P-value = the area under the 473 df t curve to the left of –2.718 ≈ 0.0034.
Since the P-value is much smaller than α, the null hypothesis of no difference is rejected.
The data provide very strong evidence to conclude that the mean salary for female MBA
graduates from this business school is lower than that for the male MBA graduates.
11.67 Let μ1 denote the true mean approval rating for male players and μ2 the true mean approval rating
for female players.
Ho: μ1 − μ 2 = 0 Ha: μ1 − μ 2 > 0
α = 0.01
( x1 − x2 ) − 0
Test statistic: t=
2 2
s1 s 2
+
n1 n 2
Assumptions: The sample sizes for each group is large (greater than or equal to 30) and the two
n1 = 56, x 1 = 2.76, s1 = 0.44, n 2 = 67, x 2 = 2.02, s2 = 0.41
(2.76 − 2.02) − 0 0.74

t= = = 9.58
2
(0.44) (0.41) 2 0.0772
+
56 67
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
⎝ n1 n2 ⎠ (0.003457 + 0.002509)2
df = 2 2
= = 113.8
1 ⎛ s12 ⎞ 1 ⎛ s22 ⎞ (0.003457)2 (0.002509)2
⎜⎜ ⎟⎟ + ⎜ ⎟ +
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎜⎝ n2 ⎟⎠ 55 66
P-value = area under the 113 df t curve to the right of 9.58 ≈ 1 − 1 = 0.
Since the P-value is less than α, the null hypothesis is rejected. At level of significance 0.05, the data
supports the conclusion that the mean approval rating is higher for males than for females.
11.68 Let μ d denote the mean difference in number of words recalled. (one-hour minus twenty-four hours)
Ho: μ d = 3 Ha: μ d > 3
α = 0.01
xd − 3
sd
n
The differences are: 4, 8, 4, 1, 2, 3, 4, 3. From these: x d = 3.625 and sd = 2.066
307
3.625 − 3
t= = .86
2.066
8
Since the P-value is greater than α, the null hypothesis is not rejected. There is not sufficient
evidence to support the conclusion that the mean number of words recalled after 1 hour exceeds the
mean number of words recalled after 24 hours by more than 3.
11.69 a. Even though the data are paired the two responses from the same subject, one in 1994 and
one in 1995, are most likely not strongly “correlated”. The questionnaire asked about alcohol
consumption during the “previous week” but the alcohol consumption pattern may vary quite
a bit within the same individual from one week to the next which would explain the low
correlation between the paired responses.
b. Let μd denote the true average difference in the number of drinks consumed by this
population between 1994 and 1995 (average for 1994 – average for 1995).
A 95% confidence interval for μd is
sd ⎛ 5.52 ⎞
d ± (t critical) ⇒ 0.38 ± (1.985) ⎜ ⎟
n ⎝ 96 ⎠
⇒ 0.38 ± 1.1185 ⇒ ( − 0.738, 1.498).
Since zero is included in the confidence interval, zero is a plausible value for μd and hence
the data do not provide evidence indicating a decrease in the mean number of drinks
consumed.
c. Let μ d denote the mean difference in number of drinks consumed by non credit card
shoppers between 1994 and 1995.
Ho: μ d = 0 Ha: μ d ≠ 0
We will compute a P-value for this test.
xd − 0
sd
n
d = 0.12 and s d = 4.58
0.12 − 0
t= = 0.764
4.58
850
P-value =2 times (the area under the 849 df t curve to the right of 0.764) ≈ 0.445.
Thus, the null hypothesis cannot be rejected at any of the commonly used significance levels
(e.g., α = 0.01, 0.05, or 0.10). There is not sufficient evidence to support the conclusion that
the mean number of drinks consumed by the non credit card shoppers has changed between
1994 and 1995.
308
11.70 Let μ d denote the true average difference in number of seeds detected by the two methods
(Direct − Stratified).
Ho: μ d = 0 (no difference in average number of seeds detected)

Ha: μ d =/ 0 (average number of seeds detected by the Direct method is not the same as the
average number of seeds detected by the Stratified method)
α = 0.05
xd − 0
sd
n
The differences are: 16, −4, −8, 4, −32, 0, 12, 0, 4, −8, 4, 12, 8, −28, 4, 0, 0, 4,
0, −8, −8, 0, 0, −4, −28, 4, −36.
From these: x d = −3.407 and s d = 13.253
−3.407 − 0
t= = −1.34
13.253
27
P-value = 2(area under the 26 df t curve to the left of −1.34) ≈ 2(0.096) = 0.192.
Since the P-value exceeds α, the null hypothesis is not rejected. The data do not provide
sufficient evidence to conclude that the mean number of seeds detected differs for the two methods.
11.71 Let μ d denote the true mean weight change.
Ho: μ d = 0 Ha: μ d > 0
α = 0.05
xd − 0
The test statistic is: z =
sd
n
5.15 − 0
z= = 8.07
11.45
322
P-value = area under the z curve to the right of 8.07 = 0.
Since the P-value is less than α, Ho is rejected. The data very strongly suggests that the true
mean weight change is positive for those who quit smoking.
11.72 Let π1 denote the proportion of elementary school teachers who are very satisfied and π2 denote the
proportion of high school teachers who are very satisfied.
Ho: π 1 − π 2 = 0 Ha: π 1 − π 2 =/ 0
α = 0.05
309
p1 − p 2
z=
p c(1 − p c ) p c(1 − p c )
+
n1 n2
224 126
p1 = = 0.567089 p2 = = 0.473684
395 266
n1p1 + n2 p2 224 + 126

pc = = = 0.529501
n1 + n2 395 + 266
(0.567089 − 0.473684) 0.093404

z= = = 2.36
0.529501(0.470499) 0.529501(0.470499) 0.039589
+
395 266
P-value = 2(area under the z curve to the right of 2.36) = 2(1 − 0.9909) = 0.0182.
Since the P-value is less than α, Ho is rejected. The data supports the claim that the proportion of
teachers who are “very satisfied” is different for elementary-school teachers than for high-school
teachers.
11.73 Let π1 denote the proportion of female Indian False Vampire bats that spend over five minutes in the
air before locating food. Let π2 denote the proportion of male Indian False Vampire bats that spend
over five minutes in the air before locating food.
Ho: π 1 − π 2 = 0 Ha: π 1 − π 2 =/ 0
α = 0.01
p1 − p 2
z=
p c(1 − p c ) p c(1 − p c )
+
n1 n2
36 64
p1 = = 0.1865, p2 = = 0.3810,
193 168
n1p1 + n2 p2 36 + 64
pc = = = 0.277
n1 + n2 193 + 168
(0.1865 − 0.3810) −0.1945

z= = = −4.12
0.277(0.723) 0.277(0.723) 0.0472
+
193 168
P-value = 2(area under the z curve to the left of -4.12) ≈ 2(0) = 0.
Since the P-value is less than α, Ho is rejected. There is sufficient evidence in the data to support the
conclusion that the proportion of female Indian False Vampire bats who spend over five minutes in
the air before locating food differs from that of male Indian False Vampire bats.
310
11.74 Let π1 denote the proportion of females who are concerned about getting AIDS and let π2 denote the
proportion of males who are similarly concerned.
Ho: π 1 − π 2 = 0 Ha: π 1 − π 2 > 0
We will compute P-value for this test.
p1 − p 2
z=
p c(1 − p c ) p c(1 − p c )
+
n1 n2
p1 = 0.427, p2 = 0.275,
n1p1 + n2 p2 568(0.427) + 234(0.275)

pc = = = 0.3827
n1 + n2 568 + 234
(0.427 − 0.275) 0.152

z= = = 4.026
0.3827(1 − 0.3827) 0.3827(1 − 0.3827) 0.0378
+
568 234
P-value = area under the z curve to the right of 4.026 ≈ 0.00003.
Since the P-value is much smaller than any of the commonly used significance values, Ho is rejected.
There is sufficient evidence in the data to support the conclusion that the proportion of females who
are concerned about getting AIDS is greater than the proportion of males so concerned.
11.75 Let π1 denote the proportion of students in the College of Computing who lose their HOPE
scholarship at the end of the first year and let π2 denote the proportion of students in the Ivan Allen
College who lose their HOPE scholarship at the end of the first year.
Ho: π 1 − π 2 = 0 Ha: π 1 − π 2 ≠ 0
We will compute a P-value for this test.
p1 − p 2
z=
p c(1 − p c ) p c(1 − p c )
+
n1 n2
p1 = 0.532, p2 = 0.649,
n1p1 + n2 p2 137(0.532) + 111(0.649)

pc = = = 0.5842
n1 + n2 137 + 111
(0.532 − 0.649) −0.11665

z= = = −1.853
0.5842(1 − 0.5842) 0.5842(1 − 0.5842) 0.06294
+
137 111
P-value =2 times (the area under the z curve to the left of –1.853) ≈ 0.0638.
311
Ho cannot be rejected at a significance level of 0.05 or smaller. There is not sufficient evidence in the
data to support the conclusion that the proportion of students in the College of Computing who lose
their HOPE scholarship at the end of one year is different from the proportion for the Ivan Allen
College.
11.76 Let π1 denote the proportion of priests in 1985 who agreed that celibacy should be a matter of
personal choice. Let π2 denote the proportion of priests in 1993 who agreed that celibacy should be a
matter of personal choice.
Ho: π 1 − π 2 = 0 Ha: π 1 − π 2 > 0
α = 0.05
p1 − p 2
z=
p c(1 − p c ) p c(1 − p c )
+
n1 n2
n1 = 200, p1 = 0.69, n 2 = 200, p2 = 0.38
n1p1 + n2 p2 200(0.69) + 200(0.38)

pc = = = 0.535
n1 + n2 200 + 200
(0.69 − 0.38) 0.31
z= = = 6.21
0.535(0.465) 0.535(0.465) 0.049877
+
200 200
P-value = area under the z curve to the right of 6.21 ≈ 0.
proportion of priests who agree that celibacy should be a matter of personal choice has declined from
1985 to 1993.
11.77 Let π1 denote the true proportion of returning students who do not take an orientation course and π2
denote the true proportion of returning students who do take an orientation course.
50 56
n1 = 94, x1 = 50, p1 = = 0.5319, n2 = 94, x 2 = 56, p2 = = 0.5957
94 94
0.5319(0.4681) 0.5957(0.4043)
(0.5319 − 0.5957) ± 1.96 + ⇒ − 0.0638 ± 1.96(0.0722)
94 94
⇒ − 0.0638 ± 0.1415 ⇒ ( − 0.2053, 0.0777).
With 95% confidence, it is estimated that the difference between the proportion of returning who do
not take an orientation course and the proportion of returning students who do take an orientation
course may be as small as −0.2053 to as large as 0.0777.
312
11.78 Let π1 denote the true proportion of students using marijuana who were exposed to the DARE
program and π2 denote the true proportion of students using marijuana who were not exposed to the
DARE program.
Ho: π 1 − π 2 = 0 Ha: π 1 − π 2 < 0
α = 0.05
p1 − p 2
z=
p c(1 − p c ) p c(1 − p c )
+
n1 n2
141 181
n1 = 288, x1 = 141, p1 = = 0.489583, n2 = 335, x 2 = 181, p2 = = 0.540299
288 335
n1p1 + n2 p2 141 + 181

pc = = = 0.516854
n1 + n2 288 + 335
(0.489583 − 0.540299) −0.050715

z= = = −1.263
0.516854(0.483146) 0.516854(0.483146) 0.040156
+
288 335
P-value = area under the z curve to the left of −1.263 = 0.1033.
Since the P-value exceeds α, Ho is not rejected. The sample data does not support the conclusion
that the proportion of students using marijuana is lower for those exposed to DARE than for those not
exposed to DARE.
11.79 Let π1 denote the proportion of games where a player suffers a sliding injury when stationary bases
are used. Let π2 denote the proportion of games where a player suffers a sliding injury when break-
away bases are used.
Ho: π 1 − π 2 = 0 Ha: π 1 − π 2 > 0
α = 0.01
p1 − p 2
z=
p c(1 − p c ) p c(1 − p c )
+
n1 n2
90 20
p1 = = 0.072, p2 = = 0.016,
1250 1250
n1p1 + n2 p2 90 + 20
pc = = = 0.044
n1 + n2 1250 + 1250
(0.072 − 0.016) 0.056

z= = = 6.83
0.044(0.956) 0.044(0.956) 0.0082
+
1250 1250
313
P-value = area under the z curve to the right of 6.83 ≈ 0.
Since the P-value is less than α, Ho is rejected. The data suggests that the use of break-away bases
reduces the proportion of games in which a player suffers a sliding injury.
11.80 Let π1 denote the true proportion of children drinking fluoridated water who have decayed teeth, and
let π2 denote the true proportion of children drinking non-fluoridated water who have decayed teeth.
67 106
n1 = 119 x1 = 67 p1 = = 0.5630 n2 = 143 x 2 = 106 p2 = = 0.7413
119 143
0.5630(0.4370) 0.7413(0.2587)
(0.5630 − 0.7413) ± 1.645 + ⇒ − 0.1783 ± 1.645(0.0584)
119 143
⇒ − 0.1783 ± 0.096 ⇒ ( − 0.2743, -0.0823).
The interval does not contain 0, so we can conclude that the two true proportions differ. Since both
endpoints of the interval are negative, this indicates that π1 < π2. Thus with 90% confidence, it is
estimated that the percentage of children drinking fluoridated water that have decayed teeth is less
than that for children drinking non-fluoridated water by as little as about 8%, to as much as 27%.
11.81 Let μ1 denote the mean number of goals scored per game for games in which Gretzky played and μ2
the mean number of goals scored per game for games in which he did not play.
Ho: μ1 − μ 2 = 0 Ha: μ1 − μ 2 > 0
α = 0.01
( x1 − x2 ) − 0
Test statistic: t=
s12 s22
+
n1 n2
n1 = 41, x 1 = 4.73 , s1 = 1.29, n 2 = 17, x 2 = 3.88, s 2 = 1.18
(4.73 − 3.88) − 0 0.85

t= = = 2.4286
2
(1.29) (1.18) 2 0.3500
+
41 17
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
⎝ n1 n2 ⎠ (0.040588 + 0.081906)2
df = 2 2
= = 32.6
1 ⎛ s12 ⎞ 1 ⎛ s22 ⎞ (0.040588)2 (0.081906)2
⎜⎜ ⎟⎟ + ⎜ ⎟ +
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎜⎝ n2 ⎟⎠ 40 16
314
Since the P-value exceeds α, Ho is not rejected. At a significance level of 0.01, the sample data does
not support the conclusion that the mean number of goals scored per game is larger when Gretzky
played than when he didn’t play.
11.82 a. Let μ1 denote the true average peak loudness for open-mouth chewing and μ2 the true
average peak loudness for closed mouth chewing. Then μ1 − μ2 denotes the difference
between the means of open-mouthed and closed-mouth chewing.
n1 = 10, x1 = 63 , s1 = 13, n 2 = 10, x2 = 54, s2 = 16
s12 (13)2 s22 (16)2

V1 = = = 16.9 V2 = = = 25.6
n1 10 n2 10
(V1 + V2 )2 (16.9 + 25.6)2 1806.25

df = = = = 17.276
V12 V22 (16.9) 2
(25.6) 2 104.552222
+ +
n1 − 1 n2 − 1 9 9
Use df = 17.
The 95% confidence interval for μ1 − μ 2 based on this sample is
(63 − 54) ± 2.11 16.9 + 25.6 ⇒ 9 ± 2.11(6.519202) ⇒ 9 ± 13.75 ⇒ ( − 4.75,22.75) .
Observe that the interval includes 0, and so 0 is one of the plausible values of μ1 − μ 2 . That
is, it is plausible that there is no difference in the mean loudness for open-mouth and closed-
mouth chewing of potato chips.
b. Let μ1 denote the true average peak loudness for closed-mouth chewing of potato chips and
μ2 the true average peak loudness for closed-mouth chewing of tortilla chips.
Ho: μ1 − μ 2 = 0 Ha: μ1 − μ 2 =/ 0
α = 0.01
( x1 − x2 ) − 0
Test statistic: t=
s12 s22
+
n1 n2
n1 = 10, x1 = 54 , s1 = 16, n2 = 10, x2 = 53, s2 = 16
(54 − 53) − 0 1.0

t= = = 0.1398
2
(16) (16) 2 7.1554
+
10 10
315
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
⎝ n1 n2 ⎠ (25.6 + 25.6)2
df = 2 2
= = 18
1 ⎛ s12 ⎞ 1 ⎛ s22 ⎞ (25.6)2 (25.6)2
⎜⎜ ⎟⎟ + ⎜ ⎟ +
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎜⎝ n2 ⎟⎠ 9 9
P-value = 2(area under the 18 df t curve to the right of 0.1398) ≈ 2(0.445) = 0.890.
Since the P-value exceeds α, Ho is not rejected. There is not sufficient evidence to conclude
that there is a difference in the mean peak loudness for closed-mouth chewing of tortilla chips
and potato chips.
c. Let μ1 denote the true average peak loudness for fresh tortilla chips when chewing closed-
mouth. Let μ2 denote the true average peak loudness of stale tortilla chips when chewing
closed-mouth.
Ho: μ1 − μ 2 = 0 Ha: μ1 − μ 2 > 0
α = 0.05
( x1 − x2 ) − 0
Test statistic: t=
s12 s22
+
n1 n2
n1 = 10, x1 = 56 , s1 = 14, n2 = 10, x2 = 53, s2 = 16
(56 − 53) − 0 3.0

t= = = 0.4462
2
(14) (16) 2 6.723
+
10 10
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
⎝ n1 n2 ⎠ (19.6 + 25.6)2
df = 2 2
= = 17.69
1 ⎛ s12 ⎞ 1 ⎛ s22 ⎞ (19.6)2 (25.6)2
⎜ ⎟ + ⎜ ⎟ +
n1 − 1 ⎜⎝ n1 ⎟⎠ n2 − 1 ⎜⎝ n2 ⎟⎠ 9 9
that there is a difference in the mean peak loudness when chewing fresh or stale tortilla chips
closed-mouth.
316
11.83 Let μ1 denote the mean number of imitations for infants who watch a human model. Let μ2 denote
the mean number of imitations for infants who watch a doll.
Ho: μ1 − μ 2 = 0 Ha: μ1 − μ 2 > 0

α = 0.01
( x1 − x2 ) − 0
Test statistic: t=
s12 s22
+
n1 n2
n1 = 12, x1 = 5.14 , s1 = 1.6, n2 = 15, x2 = 3.46, s2 = 1.3
(5.14 − 3.46) − 0 1.68

t= = = 2.94
2
(1.6) (1.3) 2 0.570964
+
12 15
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
⎝ n1 n2 ⎠ (0.1024 + 0.0676)2
df = 2 2
= = 21.1
1 ⎛ s12 ⎞ 1 ⎛ s22 ⎞ (0.1024)2 (0.0676)2
⎜ ⎟ + ⎜ ⎟ +
n1 − 1 ⎜⎝ n1 ⎟⎠ n2 − 1 ⎜⎝ n2 ⎟⎠ 11 14
The P-value is less than 0.01 and so the null hypothesis is rejected. The data supports the
conclusion that the mean number of imitations by infants who watch a human model is larger than the
mean number of imitations by infants who watch a doll.
11.84 a. Let μ d denote the mean difference in blood pressure (dental setting minus medical setting).
Ho: μ d = 0 Ha: μ d > 0
α = 0.01
xd − 0
sd
n
4.47 − 0
t= = 3.95
8.77
60
Since the P-value is less than α, Ho is rejected. Thus, the data does suggest that true mean
blood pressure is higher in a dental setting than in a medical setting.
317
b. Let μ d denote the true mean difference in pulse rate (dental minus medical).
Ho: μ d = 0 Ha: μ d =/ 0
α = 0.05
xd − 0
sd
n
−1.33 − 0
t= = − 1.165
8.84
60
P-value = 2(area under the 59 df t curve to the left of −1.165) = 2(0.124) = 0.248.
that mean pulse rates differ for a dental setting and a medical setting.
11.85 Let μ1 and μ2 denote the mean score of teenage boys and teenage girls, respectively, on the "worries
scale".
Ho: μ1 − μ 2 = 0 Ha: μ1 − μ 2 > 0
α = 0.05
( x1 − x2 ) − 0
Test statistic: t=
2 2
s1 s 2
+
n1 n 2
n1 = 78, x 1 = 67.59, s1 = 9.7, n 2 = 108, x 2 = 62.05, s2 = 9.5
(67.59 − 62.05) − 0 5.54

t= = = 3.88
2
(9.7) (9.5) 2 1.429
+
78 108
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
⎝ n1 n2 ⎠ (1.20628 + 0.83565)2
df = 2 2
= = 163.999
1 ⎛ s12 ⎞ 1 ⎛ s22 ⎞ (1.20628)2 (0.83565)2
⎜⎜ ⎟⎟ + ⎜ ⎟ +
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎜⎝ n2 ⎟⎠ 77 107
Since the P-value is less than α, Ho is rejected. It can be concluded that the mean score of teenage
boys is larger than the mean score of teenage girls on the "worries scale".
318
11.86 a. Let μ1 denote the true mean self-esteem score for students hired by the university as RA’s
and let μ2 denote the true mean self-esteem score for those not hired as RA’s.
Ho: μ1 − μ 2 = 0 Ha: μ1 − μ 2 ≠ 0
α = 0.05
( x1 − x2 ) − 0
Test statistic: t=
2 2
s1 s 2
+
n1 n 2
n1 = 69, x 1 = 83.28, s1 = 12.21, n 2 = 47, x 2 = 81.96, s2 = 12.78
(83.28 − 81.96) − 0 1.32

t= = = 0.56
2
(12.21) (12.78) 2 2.373965
+
69 47
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
⎝ n1 n2 ⎠ (2.1606 + 3.4751)2
df = 2 2
= = 95.90
1 ⎛ s12 ⎞ 1 ⎛ s22 ⎞ (2.1606)2 (3.4751)2
⎜⎜ ⎟⎟ + ⎜ ⎟ +
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎜⎝ n2 ⎟⎠ 68 46
P-value = 2(area under the 95 df t curve to the right of 0.56) = 2(0.28975) = 0.5795.
Since the P-value exceeds α, the null hypothesis is not rejected. The sample data does not
support the conclusion that the mean self-esteem score for students hired as RA’s differs
from that of students not hired as RA’s.
319
b. Let μ1 denote the true mean leadership score for students hired by the university as RA’s and
let μ2 denote the true mean leadership score for those not hired as RA’s.
Ho: μ1 − μ 2 = 0 Ha: μ1 − μ 2 ≠ 0
α = 0.05
( x1 − x2 ) − 0
Test statistic: t=
2 2
s1 s 2
+
n1 n 2
n1 = 69, x 1 = 62.51, s1 = 3.05, n 2 = 47, x 2 = 62.43, s2 = 3.36
(62.51 − 62.43) − 0 0.08

t= = = 0.1306
2
(3.05) (3.36) 2 0.612391
+
69 47
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
⎝ n1 n2 ⎠ (0.13482 + 0.24020)2
df = 2 2
= = 92.4
1 ⎛ s12 ⎞ 1 ⎛ s22 ⎞ (0.13482)2 (0.24020)2
⎜⎜ ⎟⎟ + ⎜ ⎟ +
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎜⎝ n2 ⎟⎠ 68 46
Since the P-value exceeds α, the null hypothesis is not rejected. The sample data does not
support the conclusion that the mean leadership score for students hired as RA’s differs from
that of students not hired as RA’s.
c. Let μ1 denote the true mean GPA for students hired by the university as RA’s and μ2 the true
mean GPA for those not hired as RA’s.
Ho: μ1 − μ 2 = 0 Ha: μ1 − μ 2 ≠ 0
α = 0.05
( x1 − x2 ) − 0
Test statistic: t=
2 2
s1 s 2
+
n1 n 2
n1 = 69, x 1 = 2.94, s1 = 0.61, n 2 = 47, x 2 = 2.60, s2 = 0.79
(2.94 − 2.60) − 0 0.34

t= = = 2.4882
2
(0.61) (0.79) 2 0.136644
+
69 47
320
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
⎝ n1 n2 ⎠ (0.00539 + 0.01328)2
df = 2 2
= = 81.8
1 ⎛ s12 ⎞ 1 ⎛ s22 ⎞ (0.00539)2 (0.01328)2
⎜⎜ ⎟⎟ + ⎜ ⎟ +
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎜⎝ n2 ⎟⎠ 68 46
Since the P-value is less than α, the null hypothesis is rejected. The sample data supports
the conclusion that the mean GPA for students hired as RA’s differs from that of students not
hired as RA’s.
11.87 Let μ1 denote the true mean self-esteem score for students who are members of Christian groups and
μ2 the true mean self-esteem score for students who are not members of Christian groups.
Ho: μ1 − μ 2 = 0 Ha: μ1 − μ 2 ≠ 0
α = 0.01
( x1 − x2 ) − 0
Test statistic: t=
2 2
s1 s 2
+
n1 n 2
n1 = 169, x 1 = 25.08, s1 = 10, n 2 = 124, x 2 = 24.55, s2 = 8
(25.08 − 24.55) − 0 0.53

t= = = 0.5035
2
(10) (8) 2 1.052542
+
169 124
2
⎛ s12 s22 ⎞
⎜⎜ + ⎟⎟
⎝ n1 n2 ⎠ (0.59172 + 0.51613)2
df = 2 2
= = 288.8
1 ⎛ s12 ⎞ 1 ⎛ s22 ⎞ (0.59172)2 (0.51613)2
⎜⎜ ⎟⎟ + ⎜ ⎟ +
n1 − 1 ⎝ n1 ⎠ n2 − 1 ⎜⎝ n2 ⎟⎠ 168 123
Since the P-value exceeds α, the null hypothesis is not rejected. The sample data does not support
the conclusion that the mean self-esteem score for students who are members of Christian groups
differs from that of students who are not members of Christian groups.
11.88 The large sample z test for testing a difference of two proportions requires that the samples be
independent. In this survey, the same people answered the initial question and the revised question
and so the two samples are not independent. There is no procedure from this chapter that can be
used to answer the question posed.
321
11.89 Let π1 denote the true proportion of adults born deaf who remove the implants. Let π2 denote the true
proportion of adults who went deaf after learning to speak who remove the implants.
Ho: π 1 − π 2 = 0 Ha: π 1 − π 2 =/ 0
α = 0.01
p1 − p 2
z=
p c(1 − p c ) p c(1 − p c )
+
n1 n2
n1 = 250, x1 = 75, p1 = 0.3, n 2 = 250, x2 = 25, p2 = 0.1
n1p1 + n2 p2 75 + 25
pc = = = 0.2
n1 + n2 250 + 250
(0.3 − 0.1) 0.2

z= = = 5.59
0.2(0.8) 0.2(0.8) 0.03577
+
250 250
P-value = 2(area under the z curve to the right of 5.59) ≈ 0.
Since the P-value is less than α, the null hypothesis is rejected. The data does support the fact that
the true proportion who remove the implants differs in those that were born deaf from that of those
who went deaf after learning to speak.
11.90 a. Let μ d denote the true mean difference in pH level (surface pH minus subsoil pH).
The differences are: −0.23, −0.16, −0.21, 0.26, −0.18, 0.17, 0.25, −0.20.
From these: x d = −0.0375 and sd = 0.2213
The 90% confidence interval for μ d is
sd .2213
d ± (t critical ) ⇒ − .0375 ± (1.8946)
n 8
⇒ − 0.0375 ± 0.1482 ⇒ ( − .1857, .1107).
With 90% confidence, it is estimated that the mean difference in pH is between −.1857 and
0.1107.
b. The assumption made about the underlying pH distributions is that each distribution is
normal, so the distribution of differences is normal.
322
Chapter 12
12.1 a. 0.020 < P-value < 0.025
b. 0.040 < P-value < 0.045
c. 0.035 < P-value < 0.040
d. P-value < 0.001
e. P-value > 0.100
12.2 a. df = 3 and χ 2 = 6.4. From Appendix Table 8, 0.090 < P-value < 0.095. Since the
P-value exceeds α, Ho would not be rejected. The data are consistent with the hypothesis
that the sales are equally divided among the four choices of book covers.
b. df = 3 and χ 2 = 15.3. From Appendix Table 8, 0.001 < P-value < 0.005. Since the
P-value is less than α, Ho would be rejected. The data would suggest that the sales are not
equally divided among the four choices of book covers.
c. df = 5 and χ 2 = 13.7. From Appendix Table 8, 0.015 < P-value < 0.020. Since the
P-value is less than α, Ho would be rejected. The data would suggest that the sales are not
equally divided among the six choices of book covers.
12.3 a. df = 3 and χ 2 = 19.0. From Appendix Table 8, P-value < 0.001. Since the P-value is less
than α, Ho is rejected.
b. If n = 40, then it is not advisable to use the chi-square test since one of the expected cell
frequencies (cell corresponding to nut type 4) would be less than 5.
12.4 Let πi denote the true proportion of male smoker lung cancer deaths for each of the four tar levels.
H0 : π 1 = π 2 = π 3 = π 4 = 0.25
Ha : at least one of the true proportions differs from 0.25.
α = 0.05 (used for demonstration purposes)
(observed count − expected count)2

Test statistic: χ 2 = ∑
all cells
expected count
n = 103 + 378 + 563 + 150 = 1194.
Expected count for each cell = 1194(0.25) = 298.5.
The expected counts for each cell are all greater than 5. It is stated that the sample is representative
of the male smokers who die of lung cancer so we can assume that they were randomly sampled.
(103 − 298.5 )2 (378 − 298.5)2 (563−298.5)2 (150−298.5)2

χ2= + + +
298.5 298.5 298.5 298.5
= 128.04 + 21.17 + 234.37 + 73.88 = 457.46
df = 3. From Appendix Table 8, P-value < 0.001. Since the P-value is less than α, H 0 is rejected.
There is convincing evidence that the proportion of male smoker lung cancer deaths is not the same
for the four given tar level categories.
323
12.5 a. H0 : π1 = .25, π 2 = .2 π 3 = .3 π 4 = .25
Ha : H 0 is not true.

all cells
expected count
n = 1031
Expected count for each cell = 1031(0.25) = 257.75, 1031(.2) = 206.2, 1031(.3) = 309.3
and 1031(0.25) = 257.75.
The expected counts for each cell are all greater than 5. It is stated that the sample is
representative of the male smokers who smoked low tar cigarettes so we can assume that
they were randomly sampled.
(237 − 257.75 )2 (258 − 206.2)2 (320−309.3)2 (216−257.75)2

χ2= + + +
257.75 206.2 309.3 257.75
= 1.67 + 13.01 + 0.37 + 6.76 = 21.81
df = 3. From Appendix Table 8, P-value < 0.001. Since the P-value is less than α, H 0 is
rejected. There is convincing evidence that the proportion of male low tar cigarette smokers
who subsequently died of lung cancer, is not the same for the age category at which they
started smoking.
b. If 50% of all the smokers in this population start smoking between the ages of 16 to 20 and if
all five years in this age range are equally likely, there should be about 10% starting to smoke
for each of the ages 16, 17, 18, 19, 20. This is equivalent to 20% of all the smokers started
between the ages of 16 to 17 and 30% of all the smokers started between the ages of 18 to
20.
12.6 a. Let πi denote the true proportion of fatal bicycle accidents according to time of day the
accident occurred. (i = 1, 2, 3, 4). Here 1 = Midnight to 3 am, 2 = 3 am to 6 am, 3 = 6 am to
9 am, etc.
H0 : π 1 = π 2 = π 3 = π 4 = π 5 = π 6 = π 7 = π 8 = 0.125
α = 0.05
all cells
expected count
n = 715
The expected counts for each cell are all greater than 5. It is stated that the sample is a
random sample of fatal bicycle accidents in 2004.
(38 − 89.375 )2 (29 − 89.375)2 (66−89.375)2 (77−89.375)2 (99 − 89.375 )2

χ2= + + + + +
89.375 89.375 89.375 89.375 89.375
(127 − 89.375)2 (166−89.375)2 (113−89.375)2
+ +
89.375 89.375 89.375
= 29.53 + 40.78 + 6.11 + 1.71 + 1.04 + 15.84 + 65.69 + 6.24 = 166.94
324
rejected.
There is convincing evidence that the proportion of fatal bicycle accidents are not equally
likely to occur in each of the 3-hour time periods.
b. Time period Number of Accidents

midnight and noon 38+29+66+77 = 210
noon and midnight 99+127+166+113=505
Let π 1 denote the true proportion of fatal bicycle accidents occurring between midnight and
noon, and π 2 denote the true proportion of fatal bicycle accidents occurring between noon
and midnight.
1 2
H0 : π1 = , π2 =
3 3
Ha : H 0 is not true.
α = 0.05
all cells
expected count
n = 715
1 2
Expected counts: 715( ) = 238.33 and 715( ) = 476.67.
3 3
(210 − 283.33 )2 (505 − 476.67)2

χ2= + = 18.98 + 1.68 = 20.66
283.33 476.67
rejected.
There is convincing evidence against the safety officers proposal that bicycle fatalities are
twice as likely to occur between noon and midnight as during midnight to noon.
12.7 a. Let πi denote the true proportion of fatal bicycle accidents in the month they occurred in 2004.
(i = 1, 2, 3 … 11, 12, where 1 = Jan, 2 = Feb, 3 = March etc. ).
1
H0 : π 1 = π 2 = π 3 = π 4 = π 5 = …= π 10 = π 11 = π 12 = = 0.083
12
α = 0.01
all cells
expected count
n = 719
325
(38 − 59.917)2 (32 − 59.917)2 (43−59.917)2 (59−59.917)2 (78 − 59.917 )2
χ2= + + + + +
59.917 59.917 59.917 59.917 59.917
(74 − 59.917)2 (98 − 59.917)2 (85−59.917)2 (64−59.917) 2
+ + + +
59.917 59.917 59.917 59.917
(66 − 59.917)2 (42 − 59.917)2 (40−59.917)2
+ +
59.917 59.917 59.917
= 8.02 + 13.01 + 4.78 + 0.01 + 5.46 + 3.31 + 24.2 + 10.5 + 0.28 + 0.62 + 5.36 + 6.62 =
82.2
rejected. There is convincing evidence that the proportion of fatal bicycle accidents are not
equally likely to occur in each of the 12 months.
b. The proportion for each month would reflect the number of days in the month out of the 366
days in 2004. April, June, September and November have 30 days (30/366 = .082),
February has 29 (29/366 = .079) and the rest have 31 (31/366 = .085)
c. Let πi denote the true proportion of fatal bicycle accidents in the month they occurred in 2004.
(i = 1, 2, 3 … 11, 12, where 1 = Jan, 2 = Feb, 3 = March etc. ).
H0 : π 4 = π 6 = π 9 = π11 = .082 π 2 = .079 π1 = π 3 = π 5 = π 7 = π 8 = π10 = π 12 = 0.085
Ha : at least one of the true proportions differs from H 0 .
α = 0.01
all cells
expected count
n = 719
Expected count for cells 4, 6, 9 and 11: = 719(0.082) = 58.958, for cell 2: = 719(.079) =
56.801 and for cells 1, 3, 5, 7, 8, 10, and 12: = 719(.085) = 61.115
(38 − 61.115)2 (32 − 56.801)2 (43 −61.115)2 (59 −58.958)2 (78 − 61.115 )2
χ2= + + + + +
61.115 56.801 61.115 58.958 61.115
(74 − 58.958)2 (98 − 61.115)2 (85−61.115)2 (64−58.958)2
+ + + +
58.958 61.115 61.115 58.958
(66 − 61.115)2 (42 − 58.958)2 (40−61.115)2
+ +
61.115 58.958 61.115
= 8.74 + 10.83 + 5.37 + 0.00003 + 4.67 + 3.84 + 22.26 + 9.33 + 0.43 + 0.39 + 4.88 + 7.30
= 78.04
rejected. There is convincing evidence that the proportion of fatal bicycle accidents are not
equally likely to occur in each of the 12 months taking into account the number of days in the
month.
12.8 Let πi denote the true proportion of bicycle accidents resulting in death. (i = Mon, Tues, Wed etc)
1
Ho: π Su = π M = π T = π W = π Th = π F = π Sa =
7
Ha: Ho is not true.
326
α = 0.05
all cells
expected count
Computations:
Day Su M Tu W Th F Sa Total
Frequency 14 13 12 15 14 17 15 100
Expected 14.28 14.28 14.28 14.28 14.28 14.28 14.28
(14 - 14.286)2 (13 - 14.286)2 (12 - 14.286)2 (15 - 14.286)2

χ2 = + + +
14.286 14.286 14.286 14.286
(14 - 14.286)2 (17 - 14.286)2 (15 - 14.286)2
+ + + = 1.08
14.286 14.286 14.286
df = 6. From Appendix Table 8, P-value > 0.100. Since the P-value exceeds α, the null hypothesis is
not rejected. From this data, it is not reasonable to conclude that the proportion of accidents is different
for any of the days of the week.
12.9 Let πi denote the true proportion of birds choosing color i first (i = 1, 2, 3, 4). Here 1=Blue, 2=Green,
3=Yellow, and 4=Red.
Ho: π 1 = π 2 = π 3 = π 4 = 0.25
Ha: at least one of the true proportions differ from 0.25.
α = 0.01
all cells
expected count
n = 16 + 8 + 6 + 3 = 33.
(16 − 8.25)2 (8 − 8.25)2 (6 − 8.25)2 (3 − 8.25)2

χ2 = + + +
8.25 8.25 8.25 8.25
= 7.28030 + 0.00758 + 0.61364 + 3.34091 = 11.242
df = 3. From Appendix Table 8, 0.010 < P-value < 0.015. Since the P-value is greater than α, Ho is
not rejected. The data do not provide sufficient evidence indicating a color preference.
12.10 Let πi denote the true proportion of lottery ticket purchasers in age group i (i = 1, 2, 3).
Ho: π 1 = .35, π 2 = .51, π 3 = .14

Ha: at least one of the πi is different from the hypothesized
α = 0.05
all cells
expected count
n = 36 + 130 + 34 = 200.
Expected count for each cell: 18–34: 200(.35) = 70, 35-64: 200(.51) = 102, 65+: 200(.14) = 28
None of the expected counts are less than 5. The article did not specify the method of sampling but
we will assume that the 200 purchasers of the lottery tickets can be regarded as a random sample
from the population.
327
(36 − 70)2 (130 − 102)2 (34 − 28)2
χ2 = + +
70 102 28
= 16.5143 + 7.6863 + 1.2857 = 25.4863
df = 2. From Appendix Table 8, P-value < 0.001. Since the P-value is less than α, Ho is rejected.
The data provide strong evidence to conclude that one or more of the three age groups buys a
disproportionate share of lottery tickets.
12.11 Let πi denote the proportion of all returned questionnaires accompanied by cover letter i
(i = 1, 2, 3).
1 1 1
Ho: π 1 = , π 2 = , π 3 =
3 3 3
Ha: Ho is not true.
α = 0.05
all cells
expected count
n
Computations: n = 131 = 43.67
3
(48 − 43.67)2 (44 − 43.67)2 (39 − 43.67)2

χ2 = + + = 0.429 + 0.0025 + 0.499 = 0.931
43.67 43.67 43.67
not rejected. The data does not suggest that the proportions of returned questionnaires differ for the
three cover letters.
12.12 a. Let πi denote the proportion for phenotype i (i = 1, 2, 3).

Ho: π 1 = 0.25, π 2 = 0.5, π 3 = 0.25
Ha: Ho is not true.
α = 0.05
Test statistic: χ 2 = ∑ expected count
Computations: df = 2 and the computed χ 2 value is 4.63. From Appendix Table 8,

P-value > 0.1. Since the P-value is not less than α, the null hypothesis is not rejected. The
data do not contradict the researcher's theory.
b. The analysis and conclusion would remain the same. The sample size is used only to
calculate the expected cell frequencies. It has no influence on the degrees of freedom, or the
P-value. (It does improve the fit of the χ 2 distribution to the sampling distribution of the test
statistic.)
12.13 Let π i denote the proportion of phenotype i (i = 1, 2, 3, 4).

9 3 3 1
Ho: π 1 = , π 2 = , π 3 = , π 4 =
16 16 16 16
Ha: Ho is not true
α = 0.01
328
all cells
expected count
Computations:
Phenotype
1 2 3 4 Total
Frequency 926 288 293 104 1611
Expected 906.19 302.06 302.06 100.69
(926 − 906.19)2 (288 − 302.06)2 (293 − 302.06)2

χ2 = + +
906.19 302.06 302.06
= 0.433 + 0.655 + 0.278 + 0.109 = 1.47
not rejected. The data appears to be consistent with Mendel's laws.
12.14 Let π i denote the proportion of homing pigeons who prefer direction i (i = 1, 2, 3, 4, 5, 6, 7, 8).
1
Ho: π 1 = π 2 = π 3 = π 4 = π 5 = π6 = π7 = π8 =
8
Ha: Ho is not true.
α = 0.10
all cells
expected count
Computations:
Direction 1 2 3 4 5 6 7 8 Total
Frequency 12 16 17 15 13 20 17 10 120
Expected 15 15 15 15 15 15 15 15
(12 - 15)2 (16 - 15)2 (17 - 15)2 (15 - 15)2

χ2 = + + +
15 15 15 15
(13 - 15)2 (20 - 15)2 (17 - 15)2 (10 - 15)2 72
+ + + + = = 4.8
15 15 15 15 15
not rejected. The data supports the hypothesis that when homing pigeons are disoriented in a certain
manner, they exhibit no preference for any direction of flight after take-off.
Exercises 12.15 – 12.36
12.15 a. The d.f. will be (6 - 1)(3 - 1) = 10.
b. The d.f. will be (7 - 1)(3 - 1) = 12.
c. The d.f. will be (6 - 1)(4 - 1) = 15.
329
12.16 a. α = 0.10, df = (4 − 1)(5 − 1) = 12, and χ 2 = 7.2. From Appendix Table 8,
P-value > 0.10. Since the P-value exceeds α, the null hypothesis would not be rejected. The
data are consistent with the hypothesis that educational level and preferred candidate are
independent factors.
b. α = 0.05, df = (4 − 1)(4 − 1) = 9, and χ 2 = 14.5. From Appendix Table 8,

P-value > 0.100. Since the P-value exceeds α, the null hypothesis would not be rejected.
The data are consistent with the hypothesis that educational level and preferred candidate
are independent factors.
12.17 H 0 : The proportion falling in the three credit card response categories is the same for all three
years.
H a : The proportion falling in the three credit card response categories is not the same for all three
years.
α = 0.05
all cells
expected count
Observed and expected frequencies are given in the table below (expected frequencies in
parentheses)
(row total)(column total)
Expected cell count =
grand total
2004 2005 2006

40 50 40
Definitely/Probably will 130
(43.33) (43.33) (43.33)
180 190 160
Might/Might Not/Probably Not 530
(176.67) (176.67) (176.67)
781 761 801
Definitely Will Not 2343
(781) (781) (781)
1001 1001 1001 3003
The table contains the expected counts, all of which are greater than 5. Although the data was
obtained through a telephone survey, we will assume it consists of independently chosen random
samples.
(40 − 43.33)2 (50 − 43.33)2 (40 − 43.33)2 (180 − 176.67)2
χ2 = + + +
43.33 43.33 43.33 176.67
(190 − 176.67)2 (160 − 176.67)2 (781 − 781)2 (761 − 781)2 (801 − 781)2
+ + + + + = 5.204
176.67 176.67 781 781 781
df = (3 − 1)(3 − 1) = 4. From Appendix Table 8, P-value > 0.1. As the P-value is greater than α , the
null hypothesis should not be rejected. There is not enough evidence to suggest that the proportion
falling in the three credit card response categories is different for all three years.
12.18 H 0 : The response proportions to the torture question is the same for all five countries
H a : The response proportions to the torture question is not the same for all five countries
α = 0.01
330
all cells
expected count
parentheses)
grand total
parentheses)
Response
Country Never Rarely Sometimes Often Not Sure Total
Italy 600 (400) 140 (222) 140 (244) 90 (90) 30 (44) 1000
Spain 540 (400) 160 (222) 140 (244) 70 (90) 90 (44) 1000
France 400 (400) 250 (222) 200 (244) 120 (90) 30 (44) 1000
United States 360 (400) 230 (222) 270 (244) 110 (90) 30 (44) 1000
South Korea 100 (400) 330 (222) 470 (244) 60 (90) 40 (44) 1000
Total 2000 1110 1220 450 220 5000
The table contains the expected counts, all of which are greater than 5. The data consists of
independently chosen random samples.
(600 − 400)2 (140 − 222)2 (140 − 244)2 (90 − 90)2 (30 − 44)2
χ2 = + + + +
400 222 244 90 44
(540 − 400)2 (160 − 30.09)2 (140 − 244)2 (70 − 90)2 (90 − 44)2
+ + + + +
400 222 244 90 44
(400 − 400)2 (250 − 52.33)2 (200 − 244)2 (120 − 90)2 (30 − 44)2
+ + + + +
400 222 244 90 44
(360 − 400)2 (230 − 30.09)2 (270 − 244)2 (110 − 90)2 (30 − 44)2
+ + + + +
400 222 244 90 44
(100 − 400)2 (330 − 52.33)2 (470 − 244)2 (60 − 90)2 (40 − 44)2
+ + + + + = 881.326
400 222 244 90 44
df = (5 − 1)(5 − 1) = 16. From Appendix Table 8, P-value < 0.001. Because the P-value is so small,
the null hypothesis can be rejected.
It is reasonable to conclude that the response proportions are not the same for all five countries.
12.19 a. H 0 : The proportions falling into the each of the hormone use categories is the same for
women who have been diagnosed with venous thrombosis and those who have not.
H a : The proportions falling into the each of the hormone use categories is not the same for
women who have been diagnosed with venous thrombosis and those who have not.

all cells
expected count
331
parentheses)
grand total
parentheses)
None Esterified Conjugated Equine Total

Estrogen Estrogen
Venous 372 86 121 579
Thrombosis (371.57) (123.31) (84.12)
No Venous 1439 515 289 2243
Thrombosis (1439.43) (477.69) (325.88)
Total 1811 601 410 2822
independently chosen random samples from their population.
(372 − 371.57)2 (86 − 123.31)2 (121 − 84.12)2 (1439 − 1439.43)2

χ2 = + + +
371.57 123.31 84.12 1439.43
(515 − 477.69)2 (289 − 325.88)2
+ + = 34.544
477.69 325.88
df = (3 − 1)(2 − 1) = 2. From Appendix Table 8, P-value < 0.001. Because the P-value is so
small, the null hypothesis should be rejected.
It is reasonable to conclude that the proportions falling into the each of the hormone use
categories is not the same for women who have been diagnosed with venous thrombosis and
those who have not.
b. The results of part (a) could be generalized to the population of all menopausal women who
are in the large HMO in the state of Washington.
12.20 H 0 : The distribution of political affiliation is the same for all three years.
H a : The distribution of political affiliation is not the same for all three years.

all cells
expected count
parentheses)
grand total
Year
Political Affiliation 2005 2004 2003 Total
Democrat 397 (379.71) 409 (375.65) 325 (375.65) 1131

Republican 301 (343.45) 349 (339.78) 373 (339.78) 1023
Independent/unaffiliated 458 (440.47) 397 (435.76) 457 (435.76) 1312
Other 60 (52.37) 48 (51.81) 48 (51.81) 156
Total 1216 1203 1203 3622
332
independently chosen random samples of American undergraduates.
(397 − 379.71)2 (409 − 375.65)2 (325 − 375.65)2 (301 − 343.45)2 (349 − 339.78)2
χ2 = + + + +
379.71 375.65 375.65 343.45 339.78
(373 − 339.78)2 (458 − 440.47)2 (397 − 435.76)2 (457 − 435.76)2 (60 − 52.37)2
+ + + + +
339.78 440.47 435.76 435.76 52.37
(48 − 51.81)2 (48 − 51.81)2
+ + = 26.175
51.81 51.81
the null hypothesis should be rejected. It is reasonable to conclude that the distribution of political
affiliation is not the same for all three years
12.21 H 0 : The trust in the President is the same in 2005 as it was in 2002.
H a : The trust in the President is not the same in 2005 as it was in 2002.

Test statistic: χ 2 =∑all cells
expected count
parentheses)
grand total
Year
Response 2005 2002 Total
All of the Time 132 (155.48) 180 (156.52) 312
Most of the Time 337 (431.05) 528 (433.95) 865
Some of the Time 554 (473.41) 396 (476.59) 950
Never 169 (132.06) 96 (132.94) 265
Total 1192 1200 2392
(132 − 155.48)2 (180 − 156.52)2 (337 − 431.05)2 (528 − 433.95)2 (554 − 473.41)2
χ2 = + + + +
155.48 156.52 431.05 433.95 473.41
(396 − 476.59)2 (169 − 132.06)2 (96 − 132.94)2
+ + + = 95.921
476.59 132.06 132.94
the null hypothesis should be rejected. It is reasonable to conclude that the trust in the President is
not the same in 2005 as it was in 2002.
333
12.22 H 0 : Region of residence and having a credit card are independent.
H a : Region of residence and having a credit card are not independent.
all cells
expected count
α = 0.01
parentheses)
grand total
Credit Card?
Region At least one No Credit cards Total
Northeast 401 (429.85) 164 (135.15) 565
Midwest 162 (150.64) 36 (47.36) 198
South 408 (397.89) 115 (125.11) 523
West 104 (96.62) 23 (30.38) 127
Total 1075 338 1413
(401 − 429.85)2 (164 − 135.15)2 (162 − 150.64)2 (36 − 47.36)2 (408 − 397.89)2
χ2 = + + + +
429.85 135.15 150.64 47.36 397.89
(115 − 125.11)2 (104 − 96.62)2 (23 − 30.38)2
+ + + = 15.106
125.11 96.62 30.38
df = (4 − 1)(2 − 1) = 3. From Appendix Table 8, 0.001 < P-value < 0.005. Because the P-value is
smaller than α , the null hypothesis should be rejected.
It is reasonable to conclude that there is an association between region of residence for American
undergraduates and having a credit card.
12.23 H 0 : Region of residence and whether or not the student has a balance exceeding $7000 is
independent.
H a : Region of residence and whether or not the student has a balance exceeding $7000 is not
independent.
α = 0.01
all cells
expected count
parentheses)
grand total
Balance over $7000?
Region No Yes Total
Northeast 28 (87.34) 537 (477.66) 565
Midwest 162 (53.18) 182 (290.82) 344
South 42 (80.85) 481 (444.15) 523
West 9 (19.63) 118 (107.37) 127
Total 241 1318 1559
334
(28 − 87.34)2 (537 − 477.66)2 (162 − 53.18)2 (182 − 290.82)2 (42 − 80.85)2
χ2 = + + + +
87.34 477.66 53.18 290.82 80.85
(481 − 444.15)2 (9 − 19.63)2 (118 − 107.37)2
+ + + = 339.994
444.15 19.63 107.37
df = (4 − 1)(2 − 1) = 3. From Appendix Table 8, P-value < 0.001. Because the P-value is smaller
than α , the null hypothesis should be rejected.
It is reasonable to conclude that there is an association between region of residence for American
undergraduates and whether or not they have a credit card balance of over $7000.
12.24 H 0 : Children being over weight after one year after the study began and the number of sweet drinks
consumed are independent.
H a : Children being over weight after one year after the study began and the number of sweet drinks
consumed are not independent.
α = 0.05
all cells
expected count
parentheses)
grand total
Overweight?
Number of sweet drinks/Day Yes No Total
0 22(28.92) 930 (923.08) 952
1 73 (65.22) 2074 (2081.78) 2147
2 56 (52.77) 1681 (1684.23) 1737
3 or more 102 (106.09) 3390(3385.91) 3492
Total 253 8075 8328
independently chosen random samples of children aged 2 at the beginning of the study.
(22 − 28.92)2 (930 − 923.08)2 (73 − 65.22)2 (2074 − 2081.78)2 (56 − 52.77)2
χ2 = + + + +
28.92 923.08 65.22 2081.78 52.77
(1681 − 1684.23)2 (102 − 106.09)2 (3390 − 3385.91)2
+ + + = 3.03
1684.23 106.09 3385.91
df = (4 − 1)(2 − 1) = 3. From Appendix Table 8, P-value > 0.1. Because the P-value is larger
than α , the null hypothesis should not be rejected.
There is not sufficient evidence to conclude that there is an association between whether or not
children are overweight after one year and the number of sweet drink they consume.
335
12.25 H 0 : Gender and the age at which smokers began smoking are independent .
H a : Gender and the age at which smokers began smoking are not independent.

all cells
expected count
parentheses)
grand total
Gender
Age when Smoking began Male Female Total
< 16 25(17.78) 10 (17.22) 35
16-17 24 (20.83) 17 (20.17) 41
18-20 28 (30.48) 32 (29.52) 60
≥ 21 19 (26.92) 34(26.08) 53
Total 96 93 189
independently chosen random samples of smokers.
(25 − 17.78)2 (10 − 17.22)2 (24 − 20.83)2 (17 − 20.17)2 (28 − 30.48)2
χ2 = + + + +
17.78 17.22 20.83 20.17 30.48
(32 − 29.52)2 (19 − 26.92)2 (34 − 26.08)2
+ + + = 12.091
29.52 26.92 26.08
df = (4 − 1)(2 − 1) = 3. From Appendix Table 8, 0.005 < P-value < 0.010. Because the P-value is
smaller than α , the null hypothesis should be rejected.
There is sufficient evidence to conclude that there is an association between gender and the age at
which a smoker begins smoking.
12.26 H 0 : Gender and the number of vacation days taken are independent .
H a : Gender and the number of vacation days taken are not independent.

all cells
expected count
parentheses)
grand total
336
Gender
Vacation Days taken Male Female Total
None 51(50.05) 42 (42.95) 93
1-5 21 (24.76) 25 (21.24) 46
6-10 67 (78.58) 79 (67.42) 146
11-15 111 (110.34) 94(94.66) 205
16-20 71 (75.89) 70 (65.11) 141
21-25 82 (75.35) 58 (64.65) 140
More than 25 118 (106.03) 79 (90.97) 197
Total 521 447 968
independently chosen random samples.
(51 − 50.05)2 (42 − 42.95)2 (21 − 24.76)2 (25 − 21.24)2 (67 − 78.58)2
χ2 = + + + +
50.05 42.95 24.76 21.24 78.58
(79 − 67.42)2 (111 − 110.34)2 (94 − 94.66)2 (71 − 75.89)2 (70 − 65.11)2
+ + + + +
67.42 110.34 94.66 75.89 65.11
(82 − 75.35)2 (58 − 64.65)2 (118 − 106.03)2 (79 − 90.97)2
+ + + + = 9.858
75.35 64.65 106.03 90.97
df = (7 − 1)(2 − 1) = 6. From Appendix Table 8, P-value > 0.1. Because the P-value is larger
than α , the null hypothesis should not be rejected.
There is not enough sufficient evidence to conclude that there is an association between gender and
the number of vacation days taken.
It would be reasonable to generalize this conclusion to all male and female working adults in Canada.
12.27 H 0 : Age group and whether or not the individual is considered in good health are independent .
H a : Age group and whether or not the individual is considered in good health are not independent.
α = 0.01
all cells
expected count
parentheses)
grand total
Health Status
Age Good Health Poor Health Total
18-34 920(856.67) 80 (143.33) 1000
35-54 860 (856.67) 140 (143.33) 1000
55-64 790 (856.67) 210 (143.33) 1000
Total 2570 430 3000
The table contains the expected counts, all of which are greater than 5. The data consists of people
considered to be representative of American adults.
337
(920 − 856.67)2 (80 − 143.33)2 (860 − 856.67)2 (140 − 143.33)2
χ2 = + + +
856.67 143.33 856.67 143.33
(790 − 856.67)2 (210 − 143.33)2
+ + = 68.953
856.67 143.33
df = (3 − 1)(2 − 1) = 2. From Appendix Table 8, P-value < 0.001. Because the P-value is smaller
than α , the null hypothesis should be rejected.
There is sufficient evidence to conclude that there is an association between age group and whether
or not the individual is considered in good health.
12.28 Ho: City of residence and type of vehicle used most often are independent
Ha: City of residence and type of vehicle used most often are not independent
α = 0.05
all cells
expected count
parentheses)
Pleasant North
Concord
Hills San Fran.
68 83 221
Small 372
(89.06) (107.02) (175.92
63 68 106
Compact 237
(56.74) (68.18) (112.08)
88 123 142
Midsize 353
(84.51) (101.55) (166.94)
24 18 11
Large 53
(12.69) (15.25) (25.06)
243 292 480 1015
(68 − 89.06)2 (83 − 107.02)2 (221 − 175.92)2 (63 − 56.74)2

χ2 = + + +
89.06 107.02 175.92 56.74
(68 − 68.18)2 (106 − 112.08)2 (88 − 84.51)2 (123 − 101.55)2 (142 − 166.94)2 (24 − 12.69)2
+ + + + + +
68.18 112.08 84.51 101.55 166.94 12.69
(18 − 15.25)2 (11 − 25.06)2
+ + = 49.813
15.25 25.06
df = (4 − 1)(3 − 1) = 6. From Appendix Table 8, P-value < 0.001. Hence the null hypothesis is
rejected. There is enough evidence to suggest that there is an association between the city of
residence and the type of vehicle that is used most often.
12.29 a. Ho: Gender and workaholism type are independent

Ha: Gender and workaholism type are not independent
α = 0.05 (no significance level is given in the problem so we use 0.05 for illustration.)

all cells
expected count
338
parentheses)
Gender
Female Male
Work 20 41
61
enthusiasts (27.40) (33.60)
Workaholism 32 37
Workaholics 69
types (30.99) (38.01)
Enthusiastic 34 46
80
workaholics (35.93) (44.07)
Unengaged 43 52
95
workers (42.67) (52.33)
Relaxed 24 27
51
workers (22.91) (28.09)
Disenchanted 37 30
67
workers (30.09) (36.91)
190 233 423
(20 − 27.40)2 (32 − 30.99)2 (34 − 35.93)2 (43 − 42.67)2

χ2 = + + +
27.40 30.99 35.93 42.67
(24 − 22.91)2 (37 − 30.09)2 (41 − 33.60)2 (37 − 38.01)2
+ + + +
22.91 30.09 33.60 38.01
(46 − 44.07)2 (52 − 52.33)2 (27 − 28.09)2 (30 − 36.91)2
+ + + +
44.07 52.33 28.09 36.91
= 6.852
df = (6 − 1)(2 − 1) = 5. From Appendix Table 8, P-value > 0.10. Hence the null hypothesis is
not rejected. The data are consistent with the hypothesis of no association between gender
and workaholism type.
b. Another interpretation of the lack of association between gender and workaholism type is
that, for each workaholism category, the true proportion of women who belong to this
category is equal to the true proportion of men who belong to this category.
12.30 Ho: Seat location and motion sickness are independent.

Ha: Seat location and motion sickness are not independent.
α = 0.05
all cells
expected count
339
parentheses)
Nausea No nausea
58 870
Front 928
(118.85) (809.15)
Seat 166 1163
Middle 1329
location (170.21) (1158.79)
193 806
Rear 999
(127.94) (871.06)
417 2839 3256
(58 − 118.85)2 (166 − 170.21)2 (193 − 127.94)2 (870 − 809.15)2

χ2 = + + +
118.85 170.21 127.94 809.15
(1163 − 1158.79)2 (806 − 871.06)2
+ + = 73.789
1158.79 871.06
rejected. The data provide strong evidence to conclude that seat location and nausea are dependent.
12.31 Ho: There is no dependence (i.e., independent) between handgun purchase within the year prior
to death and whether or not the death was a suicide.
Ha: There is a dependence between handgun purchase within the year prior to death and
whether or not the death was a suicide.
α = 0.05 (No significance level is given in the problem. We use 0.05 for illustration.)
Test statistic: χ 2 =∑all cells
expected count
parentheses)
Suicide Not suicide
Purchased 4 12
16
Handgun (0.27) (15.73)
No handgun 63 3921
3984
purchase (66.73) (3917.27)
67 3933 4000
(4 − 0.27)2 (63 − 66.73)2 (12 − 15.73)2 (3921 − 3917.27)2

χ2 = + + + = 53.067
0.27 66.73 15.73 3917.27
rejected.
The data provide strong evidence to conclude that there is an association between handgun
purchase within the year prior to death and whether or not the death was a suicide.
NOTE: One cell has an expected count that is less than 5. The chi-square approximation is probably
not satisfactory.
340
12.32 Ho: The proportion who believe the story describes a rape is the same for each photo group.
Ha: The proportion who believe the story describes a rape is not the same for each photo group.
α = 0.01
all cells
expected count
Photo Group
Provocative Conservative No Picture
Rape 80 104 92 276
Response (99.58) (90.95) (85.47)
Not 47 12 17 76
Rape (27.42) (25.05) (23.53)
127 116 109 352
(80 − 99.58)2 (104 − 90.95)2 (92 − 85.47)2 (47 − 27.42)2

χ2 = + + + +
99.58 90.95 85.47 27.42
(12 − 25.05)2 (17 − 23.53)2
+
25.05 23.53
= 3.850 + 1.871 + 0.500 + 13.981 + 6.795 + 1.814 = 28.810.
df = (2 − 1)(3 − 1) = 2. From Appendix Table 8, 0.001 > P-value. Since the P-value is less than α,
the null hypothesis is rejected. The data support the conclusion that the proportion who believe the
story describes a rape is not the same for the three photo groups.
12.33 Ho: The proportion of correct sex identifications is the same for each nose view.
Ha: The proportion of correct sex identifications is not the same for each nose view.
α = 0.05
all cells
expected count
Nose view
Three
Front Profile
quarter
23 26 29
Correct 78
Sex ID (26) (26) (26)
Not 17 14 11
42
Correct (14) (14) (14)
40 40 40 120
(23 − 26)2 (26 − 26)2 (29 − 26)2 (17 − 14)2

χ2 = + + + +
26 26 26 14
(14 − 14)2 (11 − 14)2
+
14 14
= 0.346 + 0.000 + 0.346 + 0.643 + 0.000 + 0.643 = 1.978
df = (2 − 1)(3 − 1) = 2. From Appendix Table 8, P-value > 0.10. Since the P-value exceeds α, the
null hypothesis is not rejected. The data does not support the hypothesis that the proportions of
correct sex identifications differ for the three different nose views.
341
12.34 a.
Country Believe Don’t Believe
Great Britain 84 116 200
West Germany 64 136 200
East Germany 44 156 200
Slovenia 110 90 200
Ireland 54 146 200
Northern 66 134 200
Ireland
422 778 1200
b. Ho: The proportion who believe in fortune tellers is the same for all six countries.
Ha: The proportion who believe in fortune tellers is not he same for all six countries.
α = 0.01
all cells
expected count
422(200)
The expected count for each “believe” cell is = 70.33.
1200
778(200)
The expected count for each “don’t believe” cell is = 129.67.
1200
(84 − 70.33)2 (64 − 70.33)2 (44 − 70.33)2 (110 − 70.33)2
χ2 = + + +
70.33 70.33 70.33 70.33
(54 − 70.33)2 (66 − 70.33)2 (116 − 129.67)2 (136 − 129.67)2
+ + + +
70.33 70.33 129.67 129.67
(156 − 129.67)2 (90 − 129.67)2 (146 − 129.67)2 (134 − 129.67)2
+ + + +
129.67 129.67 129.67 129.67
= 2.656 + 0.570 + 9.859 + 22.371

+ 3.793 + 0.267 + 1.440 + 0.309
+ 5.348 + 12.135 + 2.057 + 0.145 = 60.95
df = (6 − 1)(2 − 1) = 5. From Appendix Table 8, 0.001 > P-value. Since the P-value is less
than α, the null hypothesis is rejected. The data supports the conclusion that the proportion
who believe in fortune tellers is not the same for all six countries.
12.35 Ho: Job satisfaction and teaching level are independent.

Ha: Job satisfaction and teaching level are dependent.
α = 0.05

all cells
expected count
342
Computations:
Job satisfaction
Satisfied Unsatisfied
College 74 43 117
(63.763) (53.237)
Teaching High 224 171 395
Level School (215.270) (179.730)
126 140 266
Elementary (144.967) (121.033)
Total 424 354 778
(74 − 62.763)2 (43 − 53.237)2 (224 − 215.270)2

χ2 = + +
63.763 53.237 215.270
(171 − 179.730)2 (126 − 144.967)2 (140 − 121.023)2

+ + +
179.730 144.967 121.023
= 1.644 + 1.968 + 0.354 + 0.424 + 2.482 + 2.972 = 9.844
df = (3 − 1)(2 − 1) = 2. From Appendix Table 8, 0.010 > P-value > 0.005. Since the P-value is less
than α, Ho is rejected. The data supports the conclusion that there is a dependence between job
satisfaction and teaching level.
12.36 The data given is from a study where all the homicide data was considered and so the data was from
a census not a sample. There is no need to use an inference test when the information about the
population is already known.
Exercises 12.37 – 12.47
12.37 Let πi denote the true proportion of students graduating from colleges and universities in California in
ethnic group i ( i = 1 for White, i = 2 for Black, i = 3 for Hispanic, i = 4 for Asian, and i = 5 for other).
Ho: π 1 = 0.507, π 2 = 0.066, π 3 = 0.306, π 4 = 0.108, π 5 = 0.013

Ha: at least one of the true proportions differs from the hypothesized value.
α = 0.01
all cells
expected count
n = 1000
The expected cell counts are calculated as follows:
⎛ Expected number of graduates ⎞ ⎛ proportion of graduates from ethnic group i ⎞

⎜ ⎟ = 1000 × ⎜ ⎟
⎝ for ethnic group i ⎠ ⎝ as given in the census report ⎠
Ethnic Number of graduates Population proportion Expected number of

group in the sample according to census graduates in the
report sample
White 679 0.507 507
Black 51 0.066 66
Hispanic 77 0.306 306
Asian 190 0.108 108
Other 3 0.013 13
343
(679 − 507)2 (51 − 66)2 (77 − 306)2 (190 − 108)2 (3 − 13 )2
χ2 = + + + +
507 66 306 108 13
= 58.351 + 3.409 + 171.376 + 62.259 + 7.692 = 303.09
df =4. From Appendix Table 8, P-value < 0.001. At a significance level of α = 0.01, Ho is rejected.
The data provide very strong evidence to conclude that the proportions of students graduating from
colleges and universities in California differ from the respective proportions in the population.
12.38 Let π1, π2, π3, π4 denote the true proportions of homicides occurring during Winter, Spring, Summer,
and Fall, respectively.
Ho: π 1 = π 2 = π 3 = π 4 = 0.25
Ha: Ho is not true.
α = 0.05
all cells
expected count
Computations: n = 1361
Season Winter Spring Summer Fall

Frequency 328 334 372 327
Expected 340.25 340.25 340.25 340.25
(328 - 340.25)2 (334 - 340.25)2 (372 - 340.25)2

χ2 = + +
340.25 340.25 340.25
= 0 .4410 + 0 .1148 + 2.9627 + 0 .5160 = 4.03453.
df = 3. From Appendix Table 8, P-value > 0.100. Since the P-values exceeds α, the null hypothesis is
not rejected. The data collected does not suggest that there is a difference in the proportion of
homicides occurring in the four seasons.1
12.39 a. The table below gives the row percentages for each smoking category.
< 1/wk 1/wk 2-4/wk 5-6/wk 1/day

Never
33.00 15.79 22.42 11.17 17.62
smoked
Smoked in
18.53 12.41 23.22 14.68 31.17
the past
Currently
21.58 11.90 19.36 12.19 34.97
smokes
The proportions falling into each category appear to be dissimilar. For instance, only 17.62%
of the subjects in the “never smoked” category consumed one drink per day, whereas
34.97% of those in the “currently smokes” category consume one drink per day. Similar
discrepancies are seen for other categories as well.
b. Ho: Smoking status and alcohol consumption are independent.

Ha: Smoking status and alcohol consumption are not independent.
α = 0.05 (No significance level is not given in the problem. We use 0.05 for illustration.)

all cells
expected count
344
parentheses)
Alcohol consumption
(no. of drinks)
<1/wk 1/wk 2-4/wk 5-6/wk 1/day
Never 3577 1711 2430 1211 1910
10839
Smoked (2822.22) (1520.12) (2427.33) (1372.96) (2696.37)
Smoking Smoked in 1595 1068 1999 1264 2683
8609
status the past (2241.58) (1207.37) (1927.93) (1090.49) (2141.62)
Currently 524 289 470 296 849
2428
Smokes (632.19) (340.51) (543.74) (307.55) (604.00)
5696 3068 4899 2771 5442 21876
(3577 − 2822.22)2 (1711 − 1520.12)2 (2430 − 2427.33)2 (1211 − 1372.96)2

χ2 = + + +
2822.22 1520.12 2427.33 1372.96
(1910 − 2696.37)2 (1595 − 2241.58)2 (1068 − 1207.37)2 (1999 − 1927.93)2
+ + + +
2696.37 2241.58 1207.37 1927.93
(1264 − 1090.49)2 (2683 − 2141.62)2 (524 − 632.19)2 (289 − 340.51)2
+ + + +
1090.49 2141.62 632.19 340.51
(470 − 543.74)2 (296 − 307.55)2 (849 − 604.00)2
+ + + = 980.068
543.74 307.55 604.00
df = (3 − 1)(5 − 1) = 8. From Appendix Table 8, P-value < 0.001. Hence the null hypothesis
is rejected. The data provide strong evidence to conclude that smoking status and alcohol
consumption are not independent.
c. The result of the test in part b is consistent with our observations in part a.
12.40 a. Ho: There is no dependence between smoking status and liking to do risky things.
Ha: There is a dependence between smoking status and liking to do risky things.
α = 0.05
all cells
expected count
parentheses).
Smoker Nonsmoker
Likes risky 45 46
91
things (26.33) (64.68)
Doesn’t like 36 153
189
risky things (54.68) (134.32)
81 199 280
(45 − 26.33)2 (36 − 54.68)2 (46 − 64.68)2 (153 − 134.32)2

χ2 = + + + = 27.61
26.33 54.68 64.68 134.32
345
df = (2 − 1)(2 − 1) = 1. From Appendix Table 8, P-value < 0.001. Hence the null hypothesis
is rejected. The data provide strong evidence to conclude that there is a dependence
between liking risky things and smoking status.
b. This is an observational study, so it is not reasonable to conclude that smoking causes an

increase in the desire to do risky things. On the contrary, it is quite likely that adolescents with
a liking for risky things decide to try smoking and form a habit of it. It is also possible that
liking risky things as well as desire to smoke are both influenced by environmental factors as
well as genetic factors.
12.41 Ho: Position and Role are independent.

Ha: Position and Role are not independent.
α = 0.01
all cells
expected count
parentheses)
Initiate Participate
chase in chase
Center 28 48
76
position (39.04) (36.96)
Wing 66 41
107
position (54.96) (52.04)
94 89 183
(28 − 39.04)2 (66 − 54.96)2 (48 − 36.96)2 (41 − 52.04)2

χ2 = + + + = 10.97
39.04 54.96 36.96 52.04
rejected. The data provide strong evidence to conclude that there is an association between position
and role.
For the chi-square analysis to be valid, the observations on the 183 lionesses in the sample are
assumed to be independent.
12.42 Ho: There is no dependence between age of children and parental response.
Ha: There is a dependence between age of children and parental response.
α = 0.05
all cells
expected count
Age of Children
Preteen Teen
Very 126 149 275
Effective (139.073) (135.927)
Response Somewhat 44 41 85
Effective (42.986) (42.014)
Not at all 51 26 77
Effective (38.941) (38.059)
221 216 437
346
(126 − 139.073)2 (149 − 135.927)2 (44 − 42.986)2 (41 − 42.014)2
χ2 = + + + +
139.073 135.927 42.986 42.014
(51 − 38.941)2 (26 − 38.059)2
+
38.941 38.059
= 1.2289 + 1.2573 + 0.239 + 0.0245 + 3.7344 + 3.8209 = 10.090.
df = (3 − 1)(2 − 1) = 2. From Appendix Table 8, 0.010 > P-value > 0.005. Since the P-value is less
than α, the null hypothesis is rejected. The data supports the conclusion that there is an association
between age of children and parental response.
12.43 Ho: There is no dependence between response and region of residence.

Ha: There is a dependence between response and region of residence.
α = 0.01
all cells
expected count
Response
Agree Disagree
Northeast 130 59 189
(150.35) (38.65)
Region West 146 42 188
(149.55) (38.45)
Midwest 211 52 263
(209.22) (53.78)
South 291 47 338
(268.88) (69.12)
778 200 978
(130 − 150.35)2 (59 − 38.65)2 (146 − 149.55)2 (42 − 38.45)2

χ2 = + + + +
150.35 38.65 149.55 38.45
(211 − 209.22)2 (52 − 53.78)2 (291 − 268.88)2 (47 − 69.12)2
+ + +
209.22 53.78 268.88 69.12
= 2.754 + 10.714 + 0.084 + 0.329 + 0.015 + 0.059 + 1.820 + 7.079 = 22.855
df = (4 − 1)(2 − 1) = 3. From Appendix Table 8, 0.001 > P-value.
Since the P-value is less than α, the null hypothesis is rejected.

The data supports the conclusion that there is a dependence between response and region of
residence.
347
12.44 Ho: There is no dependence between the approach used and whether or not a donation is
obtained.
Ha: There is a dependence between the approach used and whether or not a donation is
obtained.
α = 0.05 (No significance level is given in the problem. We use 0.05 for illustration.)
all cells
expected count
parentheses)
Contribution No contribution
made made
Picture of a 18 12
30
smiling child (18.33) (11.67)
Picture of an 14 16
30
unsmiling child (18.33) (11.67)
16 10
Verbal message 26
(15.89) (10.11)
Identification of 18 4
22
charity only (13.44) (8.56)
66 42 108
(18 − 18.33)2 (14 − 18.33)2 (16 − 15.89)2 (18 − 13.44)2 (12 − 11.67)2
χ2 = + + + +
18.33 18.33 15.89 13.44 11.67
(16 − 11.67)2 (10 − 10.11)2 (4 − 8.56)2
+ + + = 6.621
11.67 10.11 8.56
df = (4 − 1)(2 − 1) = 3. From Appendix Table 8, P-value > 0.05. Hence the null hypothesis is not
rejected. The data do not provide sufficient evidence to conclude that there is a dependence
between the approach used to obtain donations and whether or not a donation is successfully
obtained.
12.45 a. It is a test of homogeneity since the number of males as well as the number of females were
fixed prior to sampling.
b. Ho: Proportions of each type of offense is the same for males and females.
Ha: Ho is not true.
α = 0.05
all cells
expected count
Public
Type of crime Violent Property Drug order Total
Male 117 150 109 124 500
(91.5) (155) (138.5) (115)
Sex
Female 66 160 168 106 500
(91.5) (155) (138.5) (115)
Total 183 310 277 230 1000
348
(117 - 91.5)2 (150 - 155)2 (109 - 138.5)2
χ2 = + +
91.5 155 138.5
(124 - 115)2 (66 - 91.5)2 (160 - 155)2

+ + +
115 (91.5) 155
= 28.51
df = (4 − 1)(2 − 1) = 3. From Appendix Table 8, 0.001 > P-value. Since the P-value is less than
α, the null hypothesis is rejected. The proportions of crimes in each category are not the same
for males and females.
12.46 a. Let π i denote the true proportion of policy holders in Astrological sign group i. ( i = 1 for
Aquarius, i = 2 for Aries, …, i = 12 for Virgo).
Ho: π 1 = π 2 = π 3 = π 4 = π 5 = π 6 = π 7 = π 8 = π 9 = π 10 = π 11 = π 12 = 1/12
Ha: at least one of the true proportions differ from 1/12
α = 0.05 (A significance level is not specified in this problem. We use α = 0.05 for illustration.)

all cells
expected count
n = 35,666 + 37,926 + 38,126 + 54,906 + 37,179 + 37,354 + 37,910 + 36,677 + 34,175 +

35,352 + 37,179 + 37,718 = 460,168
Expected count for each cell = 460,168(1/12) = 38,347.3.
(35,666 − 38,347.3)2 (37,926 − 38,347.3)2 (38,126 − 38,347.3)2 (54,906 − 38,347.3)2

χ2 = + + +
38,347.3 38,347.3 38,347.3 38,347.3
(37,179 − 38,347.3)2 (37,354 − 38,347.3)2 (37,910 − 38,347.3)2 (36, 677 − 38,347.3)2
+ + + +
38,347.3 38,347.3 38,347.3 38,347.3
(34,175 − 38,347.3)2 (35,352 − 38,347.3)2 (37,179 − 38,347.3)2 (37,718 − 38,347.3)2
+ + + +
38,347.3 38,347.3 38,347.3 38,347.3
=187.48 + 4.63 + 1.28 + 7150.16 + 35.60 + 25.73 + 4.99 + 72.76 + 453.97 + 233.97
+ 35.60 + 10.33 = 8,216.48
df =11. From Appendix Table 8, P-value < 0.001. Since the P-value is less than than α, Ho
is rejected. The data provide very strong evidence to conclude that the proportions of policy
holders are not all equal for the twelve astrological signs.
b. Sign of Capricorn covers birthdates between December 22 and January 20 which is the
summer season in Australia. One possible explanation for the higher than expected
proportion of policy holders for this sign might be that more teenagers start driving during the
summer months than any other months and hence more policies are issued during this
period.
349
c. Let πi denote the true proportion of policy holders in Astrological sign group i who make
claims ( i = 1 for Aquarius, i = 2 for Aries, …………., i = 12 for Virgo).
Ho: π 1 = 35666/ 460168 = 0.077506, π 2 = 37926/ 460168 = 0.082418,

π 3 = 38126/ 460168 = 0.082852, π 4 = 54906/ 460168 = 0.119317,
π 5 = 37179/ 460168 = 0.080794, π 6 = 37354/ 460168 = 0.081175,
π 7 = 37910/ 460168 = 0.082383, π 8 = 36677/ 460168 = 0.079703,
π 9 = 34175/ 460168 = 0.074266, π 10 = 35352/ 460168 = 0.076824,
π 11 = 37179/ 460168 = 0.080794, π 12 = 37718/ 460168 = 0.081966
Ha: at least one of the true proportions differs from the hypothesized value.
α = 0.01 (A significance level is not specified in this problem. We use α = 0.01 for illustration.)

all cells
expected count
n = 1000
The required calculations for obtaining the expected cell frequencies are summarized in the
table below. The number of policy holders of the company for the different astrological signs
is given in the second column of the table. The corresponding proportions are given in the
third column. The number of policy holders in the sample making claims for the different
astrological signs is given in column 4. The corresponding expected number of claims is
given in the last column. The expected number is calculated as follows:
⎛ Expected number of claims ⎞ Number of policy holders with this sign

⎜ ⎟ = 1000 ×
⎝ for this astrological sign ⎠ Total number of policy holders
Sign Number of Proportion of Number of Expected

policy policy holders claims in the number of
holders sample claims in the
sample
Aquarius 35666 0.077506 85 77.506
Aries 37926 0.082418 83 82.418
Cancer 38126 0.082852 82 82.852
Capricorn 54906 0.119317 88 119.317
Gemini 37179 0.080794 83 80.794
Leo 37354 0.081175 83 81.175
Libra 37910 0.082383 83 82.383
Pisces 36677 0.079703 82 79.703
Sagittarius 34175 0.074266 81 74.266
Scorpio 35352 0.076824 85 76.824
Taurus 37179 0.080794 84 80.794
Virgo 37718 0.081966 81 81.966
350
(85 − 77.506)2 (83 − 82.418)2 (82 − 82.852)2 (88 − 119.317)2
χ2 = + + +
77.506 82.418 82.852 119.317
(83 − 80.794)2 (83 − 81.175)2 (83 − 82.383)2 (82 − 79.703)2
+ + + +
80.794 81.175 82.383 79.703
(81 − 74.266)2 (85 − 76.824)2 (84 − 80.794)2 (81 − 81.966)2
+ + + +
74.266 76.824 80.794 81.966
= 0.72449 + 0.00411 + 0.00877 + 8.21987 + 0.06021 + 0.04104 + 0.00462 + 0.06617 +

0.61053 + 0.87011 + 0.12718 + 0.01138 = 10.7485
df =11. From Appendix Table 8, P-value > 0.45. At a significance level of α = 0.01, Ho
cannot be rejected. It cannot be rejected even at a significance level of α = 0.05. The data do
not provide evidence to conclude that the proportions of claims are consistent with the
proportions of policy holders for the various astrological signs. However, it is worth noting
that the proportion of claims by policy holders in the sample belonging to Capricorn is much
smaller than what would be expected based on the overall proportion of policy holders with
this sign!
12.47 Ho: The ability of individuals to make correct identifications does not differ for the brands of cola.
Ha: The ability of individuals to make correct identifications differs for the brands of cola.
α = 0.05
all cells
expected count
Computations:
Number of correct identifications
0 1 2 3 or 4 Total
Coca-Cola 13 23 24 19 79
(14.3) (23.7) (23) (18)
Pepsi-Cola 12 20 26 21 79
(14.3) (23.7) (23) (18)
Royal Crown 18 28 19 14 79
(14.3) (23.7) (23) (18)
Total 43 71 69 54 237
(13 - 14.3)2 (23 - 23.7)2 (24 - 23)2 (19 - 18)2

χ2 = + + + +
14.3 23.7 23 18
(12 - 14.3)2 (20 - 23.7)2 (26 - 23)2 (21 - 18 )2
+ + + +
14.3 23.7 23 18
(18 - 14.3)2 (28 - 23.7)2 (19 - 23)2 (14 - 18)2
+ + +
14.3 23.7 23 18
= 0.12 + 0.02 + 0.04 + 0.06 + 0.38 + 0.57 + 0.39 + 0.50 + 0.94 + 0.79 + 0.70 + 0.89 = 5.4
df = (3 − 1)(4 − 1) = 6. From Appendix Table 8, P-value > 0.10. Since the P-value exceeds α, the
null hypothesis is not rejected. The data are consistent with the hypothesis that the ability of
individuals to make correct identifications does not differ for the three brands of cola.
351
Chapter 13
13.1 a. y = −5.0 + 0.017x
b. When x = 1000, y = −5 + 0.017(1000) = 12

When x = 2000, y = −5 + 0.017(2000) = 29
50
40
30
y
20
10
0 500 1000 1500 2000 2500 3000

x
c. When x = 2100, y = −5 + (0.017)(2100) = 30.7
d. 0.017
e. 0.017(100) = 1.7
f. It is stated that the community where the given regression model is valid has no small
houses. Therefore, there is no assurance that the model is adequate for predicting usage
based on house size for small houses. Consequently, it is not advisable to use the model to
predict usage for a 500 sq.ft house.
13.2 a. When x = 10, y = −0.12 + 0.095(10) = 0.83

When x = 15, y = −0.12 + 0.095(15) = 1.305
b. 0.095
c. (0.095)(−5) = −0.475
13.3 a. The mean value of serum manganese when Mn intake is 4.0 is −2 + 1.4(4) = 3.6.
The mean value of serum manganese when Mn intake is 4.5 is −2 + 1.4(4.5) = 4.3.
5 − 3.6
b. = 1.17
1.2
P(serum Mn over 5) = P(1.17 < z) = 1 − 0.8790 = 0.121.
c. The mean value of serum manganese when Mn intake is 5 is −2 + 1.4(5) = 5.

5−5 3.8 − 5
= 0, = −1
1.2 1.2
P(serum Mn over 5) = P(0 < z) = 0.5

P(serum Mn below 3.8) = P(z < −1) = 0.1587
353
13.4 a. The estimate of the change in mean fuel efficiency associated with an increase of 1
horsepower is a decrease of 0.15 mpg.
b. It is the estimate of the mean fuel efficiency in mpg of all small cars with 100 horsepower. It
is also the predicted value of the fuel efficiency of a particular small car which has 100
horsepower.
c. yˆ = 44.0 − 0.15(300) = 44 − 45 = −1. The predicted fuel efficiency is negative 1, which is an

impossible value. This situation has occurred because the fitted equation has been used to
make predictions outside its range of validity. Small cars do not have 300 horsepower
engines. The fitted equation is valid only for the range of x values for which data has been
collected.
d. Sixty-eight percent of the variability of fuel efficiency of small cars can be explained by (or
attributed to) the linear relationship between horsepower and fuel efficiency.
e. It is the magnitude of a typical deviation from the least squares line, i.e., it is the typical
difference between actual fuel efficiency and the estimated fuel efficiency of small cars with
varying horsepower.
13.5 a. The expected change in price associated with one extra square foot of space is 47. The
expected change in price associated with 100 extra square feet of space is 47(100) = 4700.
b. When x = 1800, the mean value of y is 23000 + 47(1800) = 107600.
110000 − 107600 100000 − 107600

= 0.48 = −1.52
5000 5000
P(y > 110000) = P(0.48 < z) = 1 − 0.6844 = 0.3156
P(y < 100000) = P(z < −1.52) = 0.0643
Approximately 31.56% of homes with 1800 square feet would be priced over 110,000 dollars
and about 6.43% would be priced under 100,000 dollars.
13.6 a. y = α + βx is the equation of the population regression line and ŷ = a + bx is the equation of
the least squares line (the estimated regression line).
b. The quantity b is a statistic. It is the slope of the estimated regression line. The quantity β is
a population characteristic. It is the slope of the population regression line. The quantity b is
an estimate of β.
c. α+ βx* is the true mean y value for x = x* whereas α + βx* is a population characteristic.
a + bx* is a point estimate of the mean y value when x = x* or a + bx* is a point estimate of an
individual y value to be observed when x = x*. The quantity a + bx* is a statistic.
d. σ represents the standard deviation of the random deviation e. It is the typical deviation
about the true regression line; se is an estimate of σ.
SSResid 27.890
13.7 a. r 2 = 1− = 1− = 1 − 0.3772 = 0.6228
SSTo 73.937
SSResid 27.890 27.890

b. se2 = = 1− = = 2.5355
n−2 13 − 2 11
354
s e = 2.5355 = 1.5923
The magnitude of a typical deviation of residence half-time (y) from the population regression
line is estimated to be about 1.59 hours.
c. b = 3.4307
d. ŷ = 0.0119 + 3.4307(1) = 3.4426
13.8 a.
There is a general relationship that as treadmill run time to exhaustion increases, the 20 km
ski time decreases. The plot does suggest that the simple linear regression model may be
useful.
(106.3)2
b. Sxx = 1040.95 − = 1040.95 − 1027.2445 = 13.7055
11
(106.3)(728.7)
Sxy = 7009.91 − = 7009.91 − 7041.8918 = −31.9818
11
Sxy −31.9818
b= = = −2.3335
Sxx 13.7055
728.70 106.3
y = = 66.2455 x= = 9.6636
11 11
a = y − bx = 66.2455 − ( −2.3335)(9.6636) = 66.2455 + 22.5501 = 88.7956
yˆ = 88.7956 − 2.3335 x
355
c. b = −2.3335
d. yˆ = 88.7956 − 2.3335(10) = 65.4606
e. Only with caution, since x = 15 is not within the range of x values used to compute the
estimated regression equation.
(728.70)2
f. Syy = 48390.79 − = 48390.79 − 48273.0627 = 117.7273
11
SSResid = 48390.79 − 88.7956(728.70)−(−2.3335)(7009.91)

= 48390.79 − 64705.3537 + 16357.625
= 43.0613
SSResid 43.0613
r 2 = 1− = 1− = 1 − 0.3658 = 0.6342
SSTo 117.7273
About 63.42% of the variability in the ski time values can be explained by the linear
relationship between treadmill run time and ski time.
43.0613
g. se2 = = 4.7846, se = 4.7846 = 2.1874
9
The estimated magnitude of a typical difference between actual ski time and predicted ski
time is 2.1874.
2620.57
r = 1−
2
13.9 a. = 0.883
22398.05
2620.57
b. se = = 187.184 = 13.682 with 14 d f .
14
1235.470
13.10 a. se = = 95.036 = 9.749 with 13 d f .
13
1235.47
r = 1−
2
b. = 0.951
25321.368
356
13.11 a.
There seems to be a general tendency for y to increase at a constant rate as x increases.

However, there is also quite a bit of variability in the y values.
b. Summary values are: n = 10, ∑ x = 0.688, ∑ x2 = 0.050072, ∑ y = 0.835,

∑ y2 = 0.079491, ∑ xy = 0.060861.
⎡ (0.688)(0.835) ⎤
0.060861 − ⎢ ⎥
b= ⎣ 10 ⎦ = 0.003413 = 1.246712
⎡ (0.688) ⎤
2 0.0027376
0.050072 − ⎢ ⎥
⎣ 10 ⎦
a = 0.0835 − (1.246712)(0.0688) = −0.002274
The equation of the estimated regression line is ŷ = −0.002274 + 1.246712x.
The predicted market share, when advertising share is 0.09, would be

−0.002274 + 1.246712(0.09) = 0.10993.
⎡ (0.835)2 ⎤
c. SSTo = 0.079491 − ⎢ ⎥ = 0.0097685
⎢⎣ 10 ⎥⎦
SSResid = 0.079491 − (−0.002274)(0.835) − (1.246712)(0.060861) = 0.0055135
0.0055135
r = 1− = 1 − 0.564 = 0.436
2
0.0097685
This means that 43.6% of the total variability in market share (y) can be explained by the
simple linear regression model relating market share and advertising share (x).
0.0055135
d. se = = 0.000689 = 0 .0263 with 8 d f .
8
357
Exercises 13.12 – 13.26
13.12 σ is the standard deviation of the random error term. It describes how much the points are
spread out around the true regression line. σ b is the standard deviation of the statistic b. It
describes how much the value of b varies from sample to sample. sb is an estimate of σ b based
on the sample data.
4
13.13 a. ∑ (x − x ) 2
= 250 σb =
250
= 0.253
4
b. ∑ (x − x ) 2
= 500 σb =
500
= 0.179
No, the resulting value of σ b is not half of what it was in a. However, the resulting value of
σ b2 is half of what it was in a.
c. It would require 4 observations at each x value to yield a value of σ b which is half the value
4
calculated in a. In this case ∑ (x − x )
2
= 1000 , so σ b =
1000
= 0.1265
(6.48)2
13.14 For the x values proposed in this problem, Sxx = 4.7154 − = 1.2162. For the x values used in
12
(5.82)2
Example 13.3, Sxx = 3.1804 − = 0.3577. The larger value of S xx proposed in this exercise
12
would result in a smaller value of sb and a smaller margin of error. This leads to a narrower
confidence interval and a more accurate estimate of β .
1235.47
13.15 a. se = = 95.036 = 9.7486
13
9.7486 9.7486
sb = = = 0.1537
4024.2 63.4366
b. The 95% confidence interval for β is 2.5 ± (2.16)(0.1537) ⇒ 2.5 ± 0.33 ⇒ (2.17, 2.83).
c. The interval is relatively narrow. However, whether β has been precisely estimated or not
depends on the particular application we have in mind.
561.46
r = 1−
2
13.16 a. = 0.766
2401.85
561.46
b. se = = 43.189 = 6.5719
13
⎡ (14.1)2 ⎤
Sxx = 13.92 − ⎢ ⎥ = 0.666
⎢⎣ 15 ⎥⎦
6.5719 6.5719
sb = = = 8.0529
0.666 .8161
358
⎡ (14.1)(1438.5) ⎤
c. Sxy = 1387.2 − ⎢ ⎥ = 35.01
⎣ 15 ⎦
35.01
b= = 52.567
0.666
The 90% confidence interval for β is 52.567 ± (1.77)(8.0529) ⇒ 52.567 ± 14.254

⇒ (38.313, 66.821).
⎡ (50)(16705) ⎤
13.17 a. Sxy = 44194 − ⎢ ⎥ = 2431.5
⎣ 20 ⎦
⎡ (50)2 ⎤
Sxx = 150 − ⎢ ⎥ = 25
⎣⎢ 20 ⎦⎥
2431.5
b= = 97.26, a = 835.25 − (97.26)(2.5) = 592.1
25
b. ŷ = 592.1 + 97.26(2) = 786.62. The corresponding residual is

(y − ŷ ) = 757 − 786.62 = −29.62.
c. SSResid = 14194231 − 592.1(16705) − 97.26(44194) = 4892.06
4892.06
se = = 271.781 = 16.4858
18
16.4858
sb = = 3.2972
25
The 99% confidence interval for β, the true average change in oxygen usage associated with
a one-minute increase in exercise time is
97.26 ± (2.88)(3.2972) ⇒ 97.26 ± 9.50 ⇒ (87.76, 106.76).
13.18 a. From the computer output the P-value for testing Ho: β = 0 against Ha: β =/ 0 is found to be
0.000. Hence the null hypothesis would be rejected for any reasonable α. Therefore there
does appear to be a useful linear relationship between average wage and quit rate.
(However, it is advisable to examine a scatter plot of quit rate versus wage to confirm this.
One should be aware that it is possible to obtain a statistically significant value for the slope
in a least squares fit even though a visual examination of the scatter plot might suggest that
a straight line model is inappropriate.)
b. A confidence interval for β is required. The 95% confidence interval for β based on this
sample data is
−0.34655 ± (2.16)(0.05866) ⇒ −0.034655 ± 0.12671 ⇒ (−0.161365, 0.092055)

Other confidence levels are possible.
359
13.19 a. Ho: β = 0 Ha: β ≠ 0
α = 0.05 (for illustration)

b 15
t= with df = 42 t= = 2.8302
sb 5.3
Since the P-value is less than α, Ho is rejected. The data supports the conclusion that the
simple linear regression model specifies a useful relationship between x and y. (It is
advisable to examine a scatter plot of y versus x to confirm the appropriateness of a straight
line model for these data).
b. b ± (t critical) sb ⇒ 15 ± (2.02)(5.3) ⇒ 15 ± 10.706 ⇒ (4.294, 25.706)
Based on this interval, we estimate the change in mean average SAT score associated with
an increase of $1000 in expenditures per child is between 4.294 and 25.706.
13.20 Ho: β = 0 Ha: β ≠ 0
α = 0.05
b
t= = 1.79
sb
From the Minitab output, P-value = 0.111.
Since the P-value exceeds α, Ho is not rejected. The data does not support the conclusion that the
simple linear regression model is useful. (It is advisable to examine a scatter plot of y versus x to
confirm the appropriateness of a straight line model for these data).
13.21 Summary values are: n = 10, ∑ x = 6,970 ∑ x2 = 5,693,950 ∑ y = 10,148 ∑ y2 = 12,446,748

∑ xy = 8,406,060.
We first calculate various quantities needed to answer the different parts of this problem.
⎡ (6970)(10148) ⎤
8406060 − ⎢ ⎥ 1332904
⎣ 10 ⎦ =
b= = 1.5946498217404828560
⎡ (6970)2 ⎤ 835860
5693950 − ⎢ ⎥
⎣ 10 ⎦
[1014.8 − (1.5946498217404828560)(697)]
a= = − 96.670925753116550619
10
SSResid = 12446748 − (−96.670925753116550619)(10148)

− (1.5946498217404828560)(8406060) = 23042.474
NOTE: Using the formula SSResid = ∑y 2

−a ∑ y − b ∑ xy can lead to severe roundoff errors
unless many significant digits are carried along for the intermediate calculations. The calculations for
this problem are particularly prone to roundoff errors because of the large numbers involved. This is
the reason we have given many significant digits for the slope and the intercept estimates. You may
want to try doing these calculations with fewer significant digits. You will notice a substantial loss in
accuracy in the final answer. The formula SSResid = ∑ ˆ 2 provides a more numerically stable
(y - y)
alternative. We give the calculations based on this alternative formula in the following table :
360
[1014.8 − (1.59465)(697)]
We used b = 1.59465 and a = = − 96.6711 to calculate ŷ = a + bx .
10
y ŷ = a + bx y − yˆ ( y − yˆ )2
303 301.99 1.009 1
491 477.4 13.597 184.9
659 660.79 -1.788 3.2
683 740.52 -57.52 3308.6
922 876.07 45.935 2110
1044 1083.37 -39.37 1550
1421 1306.62 114.379 13082.6
1329 1370.41 -41.407 1714.5
1481 1513.93 -32.925 1084.1
The sum of the numbers in the last column gives SSResid = 23042.5 which is accurate to the first
decimal place.
2 23042.5
se = = 2880.31
8
2880.31
2
sb = = 0.00344592, sb = 0.0587020
835860
a. The prediction equation is CHI = -96.6711 + 1.59465 Control .
Using this equation we can predict the mean response time for those suffering a closed-head
injury using the mean response time on the same task for individuals with no head injury.
b. Let β denote the expected increase in mean response time for those suffering a closed-head
injury associated with a one unit increase in mean response time for the same task for
individuals with no head injury.
Ho: β = 0 Ha: β ≠ 0
α = 0.05
b
t= with df = 8
sb
1.59465
t= = 27.1652
0.0587020
P-value = 2(area under the 8 df t curve to the right of 27.1652) ≈ 2(0) = 0.
Since the P-value is less than α, Ho is rejected. The simple linear regression model does
provide useful information for predicting mean response times for individuals with CHI and
mean response times for the same task for individuals with no head injury. (A scatter plot of y
versus x confirms that a straight line model is a reasonable model for this problem).
c. The equation CHI = 1.48 Control says that the mean response time for individuals with CHI is
proportional to the mean response time for the same task for individuals with no head injury,
and the proportionality constant is 1.48. This implies that the mean response time for
individuals with CHI is 1.48 times the mean response time for the same task for individuals
with no head injury.
361
13.22 Summary values are: n = 15, ∑ x = 609, ∑ x2 = 28037, ∑ y = 33.1, ∑ y2 = 84.45,
∑ xy = 1156.8.
⎡ (609)(33.1) ⎤
1156.8 − ⎢ ⎥ −187.06
⎣ 15 ⎦=
b= = −0.0564863
⎡ (609)2 ⎤ 3311.60
28037 − ⎢ ⎥
⎣ 15 ⎦
[33.1 − ( −0.0564863)(609)]
a= = 4.5
15
SSResid = 84.45 − (4.5)(33.1) − (−0.0564863)(1156.8) = 0.843
2 0.843
se = = 0.06485
13
0.06485
2
sb = = 0.0000196, sb = 0.004425
3311.6
The 99% confidence interval for β is −0.056 ± (3.01)(0.004425) ⇒ −0.056 ± 0.013 ⇒

(−0.069, −0.043).
13.23 a. Let β denote the expected change in sales revenue associated with a one unit increase in
advertising expenditure.
Ho: β = 0 Ha: β ≠ 0
α = 0.05
b 52.57
t= with df = 13 t= = 6.53
sb 8.05
P-value = 2(area under the 13 df t curve to the right of 6.53) ≈ 2(0) = 0.
Since the P-value is less than α, Ho is rejected. The simple linear regression model does
provide useful information for predicting sales revenue from advertising expenditures.
b. Ho: β = 40 Ha: β > 40
α = 0.01
b− 40
Test statistic: t = with df = 13
sb
(52.57 − 40)
t= = 1.56
8.05
Since the P-value exceeds α, the null hypothesis is not rejected. The data are consistent
with the hypothesis that the change in sales revenue associated with a one unit increase in
advertising expenditure does not exceed 40 thousand dollars.
362
13.24 Summary statistics are: n = 8, ∑ y = 10.15, ∑ y2 = 28.2267, ∑ x = 13592, ∑ x2 = 44681172,
∑ xy = 29659.56.
⎡ (13592)(10.15) ⎤
29659.56 − ⎢ ⎥
⎣ 8 ⎦ = .000575
b=
⎡ (13592)2 ⎤
44681172 − ⎢ ⎥
⎣ 8 ⎦
[10.15 − (0.000575)(13592)]
a= = 0 .291715
8
SSResid = 28.2267 − (0.291715)(10.15) − (0.000575)(29659.56) = 8.2115458
2 8.2115458
se = = 1.368591
6
1.368591
sb = = 0.000252
21588364
a. Let β denote the expected change in growth rate for a one unit increase in expenditure on
research and development.
Ho: β = 0 Ha: β =/ 0
α = 0.05
b
The test statistic is: t = with df = 6.
sb
0.000575
t= = 2.28
0.000252
Since the P-value exceeds α, the null hypothesis is not rejected. The data suggests that a
simple linear regression model would not provide useful information for predicting growth rate
from research and development expenditure. This is confirmed by a scatter plot of y on x
which shows a weakly linear relationship between y and x.
b. The 90% confidence interval is 0.000575 ± (1.94)(0.000252) ⇒ 0.000575 ± 0.000489

⇒ (0.000086, 0.001064).
With 90% confidence, the expected change in growth rate associated with a one unit
increase in expenditure is estimated to be as little as 0.000086 or perhaps as much as
0.001064.
363
13.25 Let β denote the average change in milk pH associated with a one unit increase in temperature.
Ho: β = 0 Ha: β < 0
α = 0.01
b
The test statistic is: t = with d.f. = 14.
sb
Computations: n = 16, ∑ x = 678, ∑ y = 104.54,
(678)(104.54)
Sxy = 4376.36 − = −53.5225
16
(678)2
Sxx = 36056 − = 7325.75
16
−53.5225
b= = −0.0073
7325.75
a = 6.53375 − (−0.0073)(42.375) = 6.8431
SSResid = 683.447 − 6.8431(104.54) − (−0.00730608)(4376.36) = 0.0177354
.0177354
se = = .001267 =.0356
14
.0356
sb = = .000416
7325.75
−0.00730608
t= = −17.5627
0.000416
Since the P-value is less than α, H0 is rejected. There is sufficient evidence in the sample to
conclude that there is a negative (inverse) linear relationship between temperature and pH. A scatter
plot of y versus x confirms this.
13.26 Let β denote the expected change in cranial capacity associated with a one unit increase in chord
length.
Ho: β = 20 Ha: β =/ 20
α = 0.05
b− 20
The test statistic is: t = with df = 5.
sb
n = 7, ∑ x = 569, ∑ x2 = 46375, ∑ y = 6310, ∑ y2 = 5764600, ∑ xy = 515660
364
⎡ (569)(6310) ⎤
515660 − ⎢ ⎥ 2747.14
⎣ 7 ⎦=
b= = 22.25694
⎡ (569) ⎤
2 123.43
46375 − ⎢ ⎥
⎣ 7 ⎦
6310 − (22.25694)(569)
a= = −907.74269
7
SSResid = 5764600 − (−907.74269)(6310) − (22.25694)(515660) = 15442.6935
2 15442.6935
se = = 3088.5387
5
2 3088.5387
sb = = 25.0226, sb = 5.002
123.43
(22.25694 − 20)
t= = 0.45
5.002
Since the P-value exceeds α, the null hypothesis is not rejected. This new experimental data does
not contradict the prior belief that β is 20.
Exercises 13.27 – 13.33
13.27
The standardized residual plot does not exhibit any unusual features.
13.28 a. Summary values are: n = 27, ∑ x =83.6, ∑ x2 = 367.5, ∑ y = 98.35, ∑ y2 =452.38,

∑ xy = 399.383.
⎡ (83.6)(98.35) ⎤
399.383 − ⎢ ⎥ 94.8618
⎣ 27 ⎦=
b= = 0.873098
⎡ (83.6) ⎤
2 108.650
367.5 − ⎢ ⎥
⎣ 27 ⎦
365
[98.35 − (0.873098)(83.6)]
a= = 0.939223
27
Hence ŷ = a + bx = 0.939223 + 0.873098 x
b. Residuals, standard deviation of residuals and the standardized residuals are given in the
table below.
Standard Standardize
min-width max-width yˆ = a + bx Residual
Row y y − yˆ deviation d
x
of residual residual
1 1.8 2.5 2.5108 -0.0108 0.654565 -0.0165
2 2.7 2.9 3.29659 -0.39659 0.65939 -0.60145
3 2 2.15 2.68542 -0.53542 0.656084 -0.81608
4 2.6 2.9 3.20928 -0.30928 0.659108 -0.46924
5 3.15 3.2 3.68948 -0.48948 0.659876 -0.74178
6 1.8 2 2.5108 -0.5108 0.654565 -0.78036
7 1.5 1.6 2.24887 -0.64887 0.6518 -0.9955
8 3.8 4.8 4.25699 0.54301 0.658322 0.82483
9 5 5.9 5.30471 0.59529 0.648356 0.91815
10 4.75 5.8 5.08644 0.71356 0.651204 1.09576
11 2.8 2.9 3.3839 -0.4839 0.659608 -0.73361
12 2.1 2.45 2.77273 -0.32273 0.656748 -0.4914
13 2.2 2.6 2.86004 -0.26004 0.657347 -0.39559
14 2.6 2.6 3.20928 -0.60928 0.659108 -0.9244
15 2.6 2.7 3.20928 -0.50928 0.659108 -0.77268
16 2.9 3.1 3.47121 -0.37121 0.659764 -0.56264
17 5.1 5.1 5.39202 -0.29202 0.647101 -0.45128
18 10.2 10.2 9.84482 0.35518 0.474789 0.74808
19 3.5 3.5 3.99507 -0.49507 0.659371 -0.75081
20 1.2 2.7 1.98694 0.71306 0.648446 1.09964
21 1.7 3 2.42349 0.57651 0.653708 0.88191
22 1.75 2.7 2.46714 0.23286 0.654145 0.35597
23 1.7 2.5 2.42349 0.07651 0.653708 0.11704
24 1.2 2.4 1.98694 0.41306 0.648446 0.637
25 1.2 4.4 1.98694 2.41306 0.648446 3.72129
26 7.5 7.5 7.48746 0.01254 0.595598 0.02106
27 4.25 4.25 4.64989 -0.39989 0.655674 -0.60989
A plot of the standardized residuals versus x values is displayed below.

4
Row 25 in data table

3 (max-width=4.40,
Standardized residual
min-width=1.20)
-1
1 2 3 4 5 6 7 8 9 10 11
min-width
366
It is clear that the data value with min-width=1.20 and max-width=4.40 is an outlier. Including
this point in the regression fit inflates the residual sum of squares and hence the estimated
precision of predictions. We now eliminate this data point and refit the regression.
Summary values are: n = 26, ∑ x =82.4, ∑ x2 = 366.06, ∑ y = 93.95, ∑ y2 =433.02,

∑ xy = 394.103.
⎡ (82.4)(93.95) ⎤
394.103 − ⎢ ⎥ 96.3533
⎣ 26 ⎦=
b= = 0.91839
⎡ (82.4)2 ⎤ 104.915
366.06 − ⎢ ⎥
⎣ 26 ⎦
[93.95 − (0.91839)(82.4)]
a= = 0.702871
26
Hence yˆ = a + bx = 0.702871 + 0.91839 x
c. The estimate of the increase in average max-width of food containers associated with a one
centimeter increase in min-width is 0.91839 cm. The intercept has no meaning in this
problem because a min-width of zero doesn’t make sense. Food containers must have a
positive value for min-width.
d. We examine a plot of the standardized residuals versus min-width (x) and a normal
probability plot of the standardized residuals. MINITAB was used to obtain these plots. The
plot of standardized residuals versus min-width shows a pattern suggestive of nonconstant
variance of the residuals. The normal probability plot suggests failure of the assumption of
normality of the residuals. Based on these plots we have reason to doubt the validity of the
assumptions of the simple linear regression model.
2
-1
1 2 3 4 5 6 7 8 9 10 11
min-width
367
2
1
-1
-2 -1 0 1 2
normal score of standardized residual
13.29 a. The assumptions required in order that the simple linear regression model be appropriate
are:
(i) The distribution of the random deviation e at any particular x value has mean value 0.
(ii) The standard deviation of e is the same for any particular value of x.
(iii) The distribution of e at any particular x value is normal.
(iv) The mean value of vigor is a linear function of stem density.
(v) The random deviations e1, e2 ,", en associated with different observations are
independent of one another.
b.
The normal probability plot appears to follow a straight line pattern (approximately). Hence
the assumption that the random deviation distribution is normal is plausible.
368
c.
There are two residuals that are relatively large. The corresponding points are enclosed in
boxes on the graph above.
d. The negative residuals appear to be associated with small x values, and the positive
residuals appear to be associated with large x values. Such a pattern is apparently the result
of the fitted regression line being influenced by the two potential outlying points. This would
cause one to question the appropriateness of using a simple linear regression model without
addressing the issue of outliers.
13.30 a. The scatter plot below suggests the use of a curvilinear model or a transformation on the data
prior to the use of a simple linear regression. The data definitely suggests that a simple
linear regression is not appropriate on the original data.
b. The standardized residual plot below suggests that the simple linear regression model is not
adequate. The same suggestions made in part (a) holds in part (b).
369
13.31 a.
The several large residuals are marked by circles. The potentially influential observations are
marked by rectangles.
b. The residuals associated with the potentially influential observations are all negative. Without
these three, there appears to be a positive trend to the standardized residual plot. The plot
suggests that the simple linear regression model might not be appropriate.
c. There does not appear to be any pattern in the plot that would suggest that it is unreasonable
to assume that the variance of y is the same at each x value.
370
13.32 a.
There is nothing in the plot to suggest that a simple linear regression model would not be
appropriate. A person might say that the graph indicates possible curvature, and perhaps
non-constant variance. However, neither of these conditions seem to be substantiated by
further analysis.
b.
The plot exhibits the same pattern as the residual plot of part a. Only the scale on the vertical
axis is different.
371
13.33
Year X Y Y-Pred Residual
1963 188.5 2.26 1.750 0.51000
1964 191.3 2.60 2.478 0.12200
1965 193.8 2.78 3.128 −0.34800
1966 195.9 3.24 3.674 −0.43400
1967 197.9 3.80 4.194 −0.39400
1968 199.9 4.47 4.714 −0.24400
1969 201.9 4.99 5.234 −0.24400
1970 203.2 5.57 5.572 −0.00200
1971 206.3 6.00 6.378 −0.37800
1972 208.2 5.89 6.872 −0.98200
1973 209.9 8.64 7.314 1.32600
The residuals are positive in 1963 and 1964, then they are negative from 1965 through 1972,
followed by a positive residual in 1973. The residuals exhibit a pattern in the plot and thus the plot
casts doubt on the appropriateness of the simple linear regression model.
Exercises 13.34 – 13.48
13.34 A confidence interval is an interval estimate of a population mean, whereas a prediction interval is an
interval of plausible values for the single future observation.
A 95% prediction interval is one for which 95% of all possible samples would yield interval limits
capturing the future observation.
13.35 If the request is for a confidence interval for β the wording would likely be "estimate the change in the
average y value associated with a one unit increase in the x variable." If the request is for a
confidence interval for α + βx* the wording would likely be "estimate the average y value when the
value of the x variable is x*."
1 (2 − 2.5)2
13.36 a. sa+b(2) = 16.486 + = 16.486 0.06 = 4.038
20 25
b. Since 3 is the same distance from x = 2.5 as 2 is, then
s a + b(2) = s a + b(3). Hence, s a + b(3) = 4.038.
372
1 (2.8 − 2.5)2
c. sa+b(2.8) = 16.486 + = 3.817
20 25
d. sa +bx * is smallest when x* = x . Hence, for this data set sa +bx * is smallest when
x* = 2.5.
13.37 a + b(.6) = −0.6678 + 1.3957(.6) = 0.16962
1 (.6 − .485)2
s a+b(.6) = 0.8123 + = 0.2817
12 .3577
The 95% confidence interval for α + β(.6) is 0.16962 ± 2.23(0.2817)

⇒ 0.16962 ± 0.62819 ⇒ (-.45857, .79781).
13.38 a. When x* = 40, y = a+bx* = 6.843345 − 0.00730608(40) = 6.5511.
1 (40 − 42.375)2
sa+b(40) = 0.0356 + = 0.008955
16 7325.75
The 95% confidence interval for α + β(40) is 6.5511 ± (2.145)(0.008955)

⇒ 6.5511 ± 0.0193 ⇒ (6.5318, 6.5704).
b. When x* = 35, a+bx* = 6.843345 − 0.00730608(35) = 6.5876.
1 (35 − 42.375)2
sa+b(35) = 0.0356 + = 0.009414
16 7325.75

⇒ 6.5876 ± 0.0281 ⇒ (6.5595, 6.6157).
c. The value 90 is outside the range of x values for which data was collected. We cannot be
sure that the straight line model applies for x values outside the range of the sample data.
Hence I wouldn’t recommend using the data to calculate a confidence interval for α + β(90).
13.39 a. The 95% prediction interval for an observation to be made when x* = 40 would be
6.5511 ± 2.15 (0.0356)2 + (0.008955)2 ⇒ 6.5511 ± 2.15(0.0367)
⇒ 6.5511 ± 0.0789 = (6.4722, 6.6300).
b. The 99% prediction interval for an observation to be made when x* = 35 would be

6.5876 ± 2.98 (0.0356)2 + (0.009414)2 ⇒ 6.5876 ± 2.98(0.0368)
⇒ 6.5876 ± 0.1097 = (6.4779, 6.6973).
c. Yes, because x* = 60 is farther from the mean value of x, which is 42.375, than is 40 or 35.
13.40 a. r2 = 0.63
b. The P-value associated with the test that Ho: β = 0 is 0.000. Since the p-value is less than
any reasonable α, Ho is rejected and the simple linear regression model appears to be useful.
A scatter plot of y versus x should be examined to confirm this.
373
c. a + b(200) = 4.7896 + 0.014388(200) = 7.6672
The 95% confidence interval is 7.6672 ± 2.13(0.347) ⇒ 7.6672 ± 0.7391

⇒ (6.9281, 8.4063).
d. The 95% prediction interval would be 7.6672 ± 2.13 2.051 + (0.347)2

⇒ 7.6672 ± 2.13(1.47357) ⇒ 7.6672 ± 3.1387 ⇒ (4.5285, 10.8059).
8.147 + 10.065
e. a + b(300) = = 9.106.
2
0.959 0.959
(t critical)sb = 10.065 − 9.106 = 0.959 ⇒ sb = = = 0.450
(t critical ) 2.13
The 99% confidence interval for a + b(300) is 9.106 ± 2.95(0.45) ⇒ 9.106 ± 1.328
⇒ (7.778, 10.434).
13.41 a. From the MINITAB output we get
Clutch Size = 133.02 + 5.92 Snout-Vent Length
b. From the MINITAB output we get sb = 1.127
c. Let β denote the mean increase in Clutch size associated with a one unit increase in Snout-
Vent length.
Ho: β = 0 Ha: β > 0
α = 0.05 (a significance level is not specified in the problem so we use 0.05 for illustration)
b
The test statistic is: t = with df. = 12.
sb
5.919
From the MINITAB output t = = 5.25
1.127
Hence the null hypothesis is rejected. The data provide strong evidence indicating that the
slope is positive.
d. The predicted value of the clutch size for a salamander with snout-vent length of 65 is
– 133.02 + 5.919 (65) = 251.715.
e. The value 105 is very much outside the range of snout-vent length values in the available
data. The validity of the estimated regression line this far away from the range of x values in
the data set is highly questionable. Therefore, calculation of a predicted value and/or a
prediction interval for the clutch size for a salamander with snout-vent length of 205 based on
available data is not recommended.
374
13.42 Summary values are: n = 27, ∑ x =83.6, ∑ x2 = 367.5, ∑ y = 98.35, ∑ y2 =452.3775,
∑ xy = 399.3825. In Exercise 13.28 we obtained b = 0.873098 and a = 0.939223 .
Hence ŷ = a + bx = 0.939223 + 0.873098 x . We will need the following quantities to answer the
various parts of this Exercise.
⎡ (83.6)2 ⎤
Sxx = 367.5 − ⎢ ⎥ = 108.650
⎢⎣ 27 ⎥⎦
SSResid = 452.3775 − 0.939223 (98.35) − 0.873098 (399.3825) = 11.3049
11.3049 0.6725
se = = 0.6725 sb = = 0.06452
25 108.650
a. Let β denote the mean increase in maximum width associated with a unit increase in the
minimum width.
Ho: β = 0 Ha: β > 0
α = 0.05 (a significance level is not specified in the problem so we use 0.05 for illustration)
b
sb
0.873098
We have t = = 13.5322
0.06452
Hence the null hypothesis is rejected. The data provide strong evidence indicating that there
is a positive linear relationship between mean minimum width and the maximum width of a
food container. Of course, this should be confirmed by examining a scatter plot of max-width
versus min-width.
b. The estimated standard deviation se was calculated above.

11.3049
We obtained s e = = 0.6725 cm.
25
This is the estimated value of the standard deviation of the distribution of values for maximum
width for food containers at any given fixed value of minimum width.
c. The mean maximum width of products with minimum width equal to 6 cm is α + β (6).
An estimate of this quantity is ŷ = 0.939223 + 0.873098(6) = 6.178 cm.
A 95% confidence interval for this quantity is
1 (6 − 3.0963)2
6.178 ± (2.06)(0.6725) + ⇒ 6.178 ±0.4691 ⇒ (5.709, 6.647).
27 108.650
d. A 95% prediction interval for the maximum width of a food package with a minimum width of
6 cm is
1 (6 − 3.0963)2
6.178 ± (2.06)(0.6725) 1 + + ⇒ 6.178 ± 1.463 ⇒ (4.715, 7.641).
27 108.650
375
13.43 a.
⎡ (269)(51) ⎤
1081.5 − ⎢ ⎥
b= ⎣ 14 ⎦ = 101.571 = 0.04462
⎡ (269) ⎤
2 2276.357
7445 − ⎢ ⎥
⎣ 14 ⎦
a = 3.6429 − (0.04462)(19.214) = 2.78551
The equation of the estimated regression line is y = 2.78551 + 0.04462x.
b. Ho: β = 0 Ha: β =/ 0
α = 0.05
b
sb
SSResid = 190.78 − (2.78551)(51) − (0.00462)(1081.5) = 0.46246
2 0.46246
se = = 0.0385
12
0.0385
2
sb = = 0 .0000169, sb = 0 .004113
2276.357
0.04462
t= = 10.85
0.004113
P-value = 2(area under the 12 df t curve to the right of 10.85) = 2(0) = 0.
Since the P-value is less than α, the null hypothesis is rejected. The data suggests that the
simple linear regression model provides useful information for predicting moisture content
from knowledge of time.
c. The point estimate of the moisture content of an individual box that has been on the shelf 30
days is 2.78551 + 0.04462(30) = 4.124.
The 95% prediction interval is
1 (30 − 19.214)2
4.124 ± (2.18) 0.0385 1+ +
14 2276.357
⇒ 4.124 ± 2.18(0.2079) ⇒ 4.124 ± 0.453 = (3.671, 4.577).
d. Since values greater than equal to 4.1 are included in the interval constructed in c, it is very
plausible that a box of cereal that has been on the shelf 30 days will not be acceptable.
13.44 Since 17 is farther away from x = 19.21 than is 20, the confidence interval with x* = 17 would be
wider. The same would be true for a prediction interval.
376
13.45 a.
⎡ (1350)(600) ⎤
57760 − ⎢ ⎥
b= ⎣ 15 ⎦ = 3760 = 0.1109
⎡ (1350) ⎤
2 33900
155400 − ⎢ ⎥
⎣ 15 ⎦
a = 40 − (0.1109)(90) = 30.019
The equation for the estimated regression line is ŷ = 30.019 + 0.1109x.
b. When x = 100, the point estimate of α + β(100) is 30.019 + 0.1109(100) = 41.109.
SSResid = 24869.33 − (30.019)(600) − (0.1109)(57760) = 452.346
2 452.346
se = = 34.7958
13
⎡ 1 (100 − 90)2 ⎤
sa2+ b(100) = 34.7958 ⎢ + ⎥ = 2.422
⎢⎣15 33900 ⎥⎦
s a+b(100) = 2.422 = 1.5564
The 90% confidence interval for the mean blood level for people who work where the air lead
level is 100 is 41.109 ± (1.77)(1.5564) ⇒ 41.109 ± 2.755 ⇒ (38.354, 43.864).
c. The prediction interval is 41.109 ± (1.77) 34.7958 + 2.422 ⇒ 41.109 ± 10.798

⇒ (30.311, 51.907).
d. The interval of part b is for the mean blood level of all people who work where the air lead
level is 100. The interval of part c is for a single randomly selected individual who works
where the air lead level is 100.
13.46 a. When x* = 35, a+bx* = 4.5 − 0.0565(35) = 2.5225
0.8430
se = = 0.25465
13
1 (35 − 40.6)2
s a+b(35) =0.25465 + =0.07026
15 3311.6

⇒ 2.5225 ± 0.1518 ⇒ (2.371, 2.674).
b. When x* = 45, a+bx* = 4.5 − 0.0565(45) = 1.9575
1 (45 − 40.6)2
s a+b(45) =0.25465 + = 0.06857
15 3311.6

⇒ 1.9575 ± 0.1481 ⇒ (1.809, 2.106).
So the interval from part a and the interval in part b form a set of simultaneous confidence
intervals with confidence level of at least 90%.
377
c. The simultaneous level would be at least [100 - 2(1)]% = 98%.
d. The simultaneous level would be at least [100 - 3(5)]% = 85%.
e. The simultaneous level would be at least [100 - 3(1)]% = 97%.
13.47 a. The 95% prediction interval for sunburn index when distance is 35 is
2.5225 ± 2.16 (0.25465)2 + (0.07026)2 ⇒ 2.5225 ± 0.5706 ⇒ (1.9519, 3.0931).
The 95% prediction interval for sunburn index when distance is 45 is
1.9575 ± 2.16 (0.25465)2 + (0.06857)2 ⇒ 1.9575 ± 0.5696 ⇒ (1.3879, 2.5271).
The pair of intervals form a set of simultaneous prediction intervals with prediction level of at
least [100 − 2(5)]% = 90%.
b. The simultaneous prediction level would be at least [100 − 3(1)]% = 97%.
13.48 Summary values are: n = 6, ∑ x = 496, ∑ x2 = 41272, ∑ y = 69.6, ∑ y2 = 942.28, ∑ xy = 5942.6
SSTo = 134.92, SSResid = 2.29
b = 0.70173, a = −46.41, se2 = 0.5725
Let α + β(82) denote the true average number of hours of chiller operation when maximum outdoor
temperature is 82.
Ho: α+ β(82) = 12 Ha: α+ β(82) < 12
α = 0.01
[a + b(82)] − 12
s a+b(82)
a + b(82) = −46.41 + 0.70173(82) = 11.132
⎡ 1 (82 − 82.667)2 ⎤
sa2+ b(82) = 0.5725 ⎢ + ⎥ = 0.0963
⎣⎢ 6 269.333 ⎦⎥
s a+b(82) = 0.3104
(11.132 − 12)
t= = −2.80
0.3104
P-value = area under the 4 df t curve to the left of −2.80 = 0.024.
Since the P-value exceeds α, the null hypothesis is not rejected. The data are consistent with the
hypothesis that the true average number of hours of chiller operation is not less than 12 when
maximum outdoor temperature is 82. The manufacturer is advised not to produce this system.
378
Exercises 13.49 – 13.57
13.49 The quantity r is a statistic as its value is calculated from the sample. It is a measure of how strongly
the sample x and y values are linearly related. The value of r is an estimate of ρ. The quantity ρ is a
population characteristic. It measures the strength of linear relationship between the x and y values
in the population.
13.50 a. Just because r = 1, it is not necessarily true that ρ = 1. The sample points might fall exactly
on a straight line even though not all population pairs do.
b. Yes. If ρ = 1, then r will be 1 for every sample, since a perfect linear relation in the population
implies that all (x,y) pairs in the sample will lie on the same line.
13.51 Let ρ denote the true correlation coefficient between teaching evaluation index and annual raise.
Ho: ρ = 0 Ha: ρ ≠ 0
α = 0.05
r
t= with df. = 351
1− r 2
n−2
n = 353, r = 0.11
0.11 0.11
t= = = 2.07
1 − (0.11) 2 0.05305
351
The t curve with 351 df is essentially the z curve.
P-value = 2(area under the z curve to the right of 2.07) = 2(0.0192) = 0.0384.
Since the P-value is less than α, Ho is rejected. There is sufficient evidence in the sample to
conclude that there appears to be a linear association between teaching evaluation index and annual
raise.
According to the guidelines given in the text book, r = 0.11 suggests only a weak linear relationship.
Since r2 = 0.0121, fitting the simple linear regression model to the data would result in only about
1.21% of observed variation in annual raise being explained.
13.52 Let ρ denote the true correlation coefficient between annual dollar rent and annual dollar sales.
Ho: ρ = 0 Ha: ρ > 0
α = 0.05
r
t= with df. = 51
1− r 2
n−2
n = 53, r = 0.37
379
0.37 0.37
t= = = 2.84
1 − (0.37)2 0.13
51
Since the P-value is less than α, Ho is rejected. The sample data supports the conclusion that there
is a positive linear association between annual dollar rent and annual dollar sales.
13.53 a. Let ρ denote the correlation coefficient between time spent watching television and grade
point average in the population from which the observations were selected.
Ho: ρ = 0 Ha: ρ < 0
α = 0.01
r
t= with df. = 526
1− r 2
n−2
n = 528, r = −0.26
−0.26 −0.26
t= = = −6.175
1 − ( −0.26) 2 0.042103
526
P-value = area under the z curve to the left −6.175 ≈ 0.
Since the P-value is less than α, Ho is rejected. The data does support the conclusion that
there is a negative correlation in the population between the two variables, time spent
watching television and grade point average.
b. The coefficient of determination measures the proportion of observed variation in grade point
average explained by the regression on time spent watching television. This value would be
(−0.26)2 = 0.0676. Thus only 6.76% of the observed variation in grade point average would
be explained by the regression. This is not a substantial percentage.
13.54 a. Let ρ denote the true correlation between particulate pollution and luminance.
Ho: ρ = 0 Ha: ρ > 0
No significance level was specified, so α = 0.01 will be used.
r
t= with df. = 13
1− r 2
n−2
⎡ (860)(348) ⎤
Sxy = 22265 − ⎢ ⎥ = 2313
⎣ 15 ⎦
380
⎡ (860)2 ⎤
Sxx = 56700 − ⎢ ⎥ = 7393.333
⎣⎢ 15 ⎦⎥
381
⎡ (348)2 ⎤
Syy = 8954 − ⎢ ⎥ = 880.4
⎣⎢ 15 ⎦⎥
2313
r = = 0.9066
(7393.333)(880.4)
0.9066
t= = 7.75
1 − (0.9066)2
13
Since the P-value is less than α, Ho is rejected. The data supports the conclusion that there
is a positive correlation between particulate pollution and luminance.
b. r2 = (0.9066)2 = 0.822. So, 82.2% of the observed variation in luminance can be attributed to
the approximate linear relationship between luminance and particulate pollution.
13.55 From the summary quantities:
⎡ (136.02)(39.35) ⎤
Sxy = 673.65 − ⎢ ⎥ = 78.94
⎣ 9 ⎦
⎡ (136.02)2 ⎤
Sxx = 3602.65 − ⎢ ⎥ = 1546.93
⎣⎢ 9 ⎦⎥
⎡ (39.35)2 ⎤
Syy = 184.27 − ⎢ ⎥ = 12.223
⎢⎣ 9 ⎥⎦
78.94 78.94
r= = =0.574
(1546.93)(12.223) 137.51
Let ρ denote the correlation between surface and subsurface concentration.
Ho: ρ = 0 Ha: ρ ≠ 0
α = 0.05
r
t= with df. = 7
1− r 2
n−2
0.574
t= = 1.855
1 − (0.574)2
7
Since the P-value exceeds α, Ho is not rejected. The data does not support the conclusion that there
is a linear relationship between surface and subsurface concentration.
382
13.56 a. With level of significance 0.001 and P-value of 0.00032, the null hypothesis would be
rejected. It would be concluded that ρ is not zero.
b. The conclusion in a is that ρ =/ 0. The small P-value does not indicate what value ρ may
have, other than that it is not zero. It does not indicate that ρ is large.
13.57 Ho: ρ = 0 Ha: ρ ≠ 0
α = 0.05
r
t= with df. = 9998
1− r 2
n−2
n = 10000, r = 0.022
0.022
t= = 2.2
1 − (0.022)2
9998
P-value = 2(area under the z curve to the right of 2.2) = 2(0.0139) = 0.0278.
Since the P-value is less than α, Ho is rejected. The results are statistically significant. Because of
the extremely large sample size, it is easy to detect a value of ρ which differs from zero by a small
amount. If ρ is very close to zero, but not zero, the practical significance of a non-zero correlation
may be of little consequence.
Exercises 13.58 – 13.75
13.58 a. b = −0.640
b. y = 106.3 − 0.64(40) = 106.3 − 25.6 = 80.7
c. Because the estimated slope is negative, the value of r is the negative square root of (0.47),
which is −0.6856.
Let ρ denote the correlation coefficient between green biomass concentration and elapsed
time since snowmelt.
Ho: ρ = 0 Ha: ρ ≠ 0
α = 0.05
r
t= with df. = 56
1− r 2
n−2
−0.6856 −0.6856
t= = = −7.05
1 − 0.47 0.0973
56
383
P-value = 2(area under the 56 df t curve to the left of −7.05) ≈ 2(0) = 0.
Since the P-value is less than α, the null hypothesis is rejected. Thus, the data does suggest
that there is a useful linear relationship between elapsed time since snowmelt and green
biomass concentration.
−0.18 −0.18
13.59 a. t= = = −3.40 with df = 345
1 − ( −.18)2 0.052959
345
If the test was a one-sided test, then the P-value equals the area under the z curve to the left
of −3.40, which is equal to 0.0003. It the test was a two-sided test, then the P-value is
2(0.0003) = 0.0006. While the researchers’ statement is true, they could have been more
precise in their statement about the P-value.
b. From my limited experience, I have observed that the more visible a person’s sense of
humor, the less depressed they appear to be. This would suggest a negative correlation
between Coping Humor Scale and Sense of Humor.
c. Since r2 = (−0.18)2 = 0.0324, only about 3.24% of the observed variability in sense of humor
can be explained by the linear regression model. This suggests that a simple linear
regression model may not give accurate predictions.
13.60 The summary values are: n = 13, x = 7, y = 36.1538,
Sxy = 577 , Sxx = 182 , Syy = 2125.6923
a. Let ρ denote the correlation between depth of flooding and flood damage.
Ho: ρ = 0 Ha: ρ > 0
α = 0.05
r
t= with df. = 11
1− r 2
(n− 2)
577
r= =0.9277
(182)(2125.6923)
0.9277
t= = 8.24
1 − (0.9277)2
11
Since the P-value is less than α, the null hypothesis is rejected. The data does suggest the
existence of a positive correlation between flood depth and damage.
b. When x = 3.5, y = 13.9617 + 3.1703(3.5) = 25.0578
SSResid = 2125.6923 − 3.1703(577) = 296.4292
384
296.4292
se2 = = 26.9481
11
⎡ 1 (3.5 − 7)2 ⎤
sa2+ b(3.5) = 26.9481⎢ + ⎥ = 3.8867
⎣⎢13 182 ⎦⎥
The 95% prediction interval for flood damage for a claim made when depth of flooding
is 3.5 feet is:
25.0578 ± 2.2 26.9481 + 3.8867 ⇒ 25.0578 ± 12.2164 ⇒ (12.8414, 37.2742).
13.61 a. Ho: β = 0 Ha: β ≠ 0
α = 0.01
b
t=
sb
From the Minitab output, t = −3.95 and the P-value = 0.003. Since the P-value is less than α,
Ho is rejected. The data supports the conclusion that the simple linear regression model is
useful.
b. A 95% confidence interval for β is −2.3335 ± 2.26(0.5911) ⇒ −2.3335 ± 1.3359

⇒ (−3.6694, −0.9976).
c. a+b(10) = 88.796 − 2.3335(10) = 65.461
s a +b (10 ) = 0.689
The 95% prediction interval for an individual y when x = 10 is
65.461 ± 2.26 (0.689)2 + 4.789 ⇒ 65.461 ± 5.185 ⇒ (60.276, 70.646).
d. Because x = 11 is farther from x than x = 10 is from x .
13.62 a. Ho: β = 0 Ha: β ≠ 0
From the Minitab output t = 15.26 and the P-value = 0. The null hypothesis is rejected and
there appears to be a useful linear relationship between length and size.
b. Ho: β ≥ 0.8 Ha: β < 0.8
α = 0.05
b − .8
t= with df. = 9
sb
0.72907 − 0.8 −0.07093

t= = = −1.48
0.04778 0.04778
P-value = area under the 9 df t curve to the left of −1.48 ≈ 0.086.
Since the P-value exceeds α, Ho is not rejected. The sample data does not support the
conclusion that the average change in maximum size associated with a 1 mm. increase in
length is less than 0.8.
385
c. a + b(325) = −89.09 + 0.72907(325) = 147.86
1 (325 − 343.27)2
sa + b(325) = 12.56 + = 3.886
11 69112.18
The 95% confidence interval is
147.86 ± 2.26(3.886) ⇒ 147.86 ± 8.78 ⇒ (139.08, 156.64).
d. Because the slope is positive, the estimated average maximum size when x=250 would be
smaller than that for x=350. Also, 250 − 343.27 = −93.27 and 325 − 343.27 = −18.27. So,
250 is farther from x than 325 is. Hence, the confidence interval for mean maximum size
when x = 250 would be wider than the interval for when x = 325.
13.63 a. Let ρ denote the correlation coefficient between soil hardness and trail length.
Ho: ρ = 0 Ha: ρ < 0
α = 0.05
r
t= with df. = 59
1− r 2
(n− 2)
−0.6213
t= = −6.09
1 − ( −0.6213)2
59
Since the P-value is less than α, the null hypothesis is rejected. The data supports the
conclusion of a negative correlation between trail length and soil hardness.
b. When x* = 6, a+b(6) = 11.607 − 1.4187(6) = 3.0948
⎡ 1 (6 − 4.5)2 ⎤
sa2+ b(6) = (2.35)2 ⎢ + ⎥ = 0.1402
⎢⎣ 61 250 ⎥⎦
sa + b(6) = 0.1402 = 0.3744
The 95% confidence interval for the mean trail length when soil hardness is 6 is
3.0948 ± 2.00(0.3744) ⇒ 3.0948 ± 0.7488 ⇒ (2.346, 3.844).
c. When x* = 10, a+b(10) = 11.607 − 1.4187(10) = −2.58
According to the least-squares line, the predicted trail length when soil hardness is 10 is
−2.58. Since trail length cannot be negative, the predicted value makes no sense. Therefore,
one would not use the simple linear regression model to predict trail length when hardness is
10.
386
13.64 a.
The plot suggests that a straight line model might adequately describe the relationship
between percent light absorption and peak photovoltage.
b.
⎡ (179.7)(7.28) ⎤
Sxy = 17.683 − ⎢ ⎥ = 33.326
⎣ 9 ⎦
⎡ (179.7)2 ⎤
Sxx = 4334.41 − ⎢ ⎥ = 746.4
⎣⎢ 9 ⎦⎥
⎡ (7.28)2 ⎤
Syy = 7.4028 − ⎢ ⎥ = 1.514
⎣⎢ 9 ⎦⎥
33.326
b= =0.044649
746.4
a = 0.809 − 0.044649(19.667) = −0.08259
y = −0.08259 + 0.044649x
2 (33.326)2
c. r = =.983
(746.4)(1.514)
d. When x* = 19.1, y = −0.08259 + 0.044649(19.1) = 0.7702.
The corresponding residual is 0.68 − 0.7702 = −0.0902.
e. Ho: β = 0 Ha: β =/ 0
α = 0.05
387
b
The test statistic is: t = with df. = 7
sb
b = 0.044649
2 7.4028 − ( −0.08259)(7.28) − 0.044649(178.683) 0.02604

se = = = 0.00372
7 7
0.00372
2
sb = = 4.984 × 10−6 , sb = 0.00223
746.4
0.044649
t= = 20.00
0.00223
P-value = 2(area under the 7 df t curve to the right of 20) ≈ 0.
Since the P-value is less than α, H0 is rejected. The data does support the conclusion that
there is a useful linear relationship between percent light absorption and peak photovoltage.
f. In the absence of a specified confidence level, 95% will be used. The 95% confidence
interval for the average change in peak photovoltage associated with a 1% increase in light
absorption is
0.044649 ± 2.37(0.00223) ⇒ 0.044649 ± 0.00529 ⇒ (0.039359, 0.049939).
g. In the absence of a specified confidence level, 95% will be used. When x* = 20,
y = −0.08259 + 0.044649(20) = 0.8104.
1 (20 − 19.667)2
s a+b(20) =0.061 + = 0.0204
9 746.4
The 95% confidence interval of true average peak photovoltage when percent light
absorption is 20 is
0.8104 ± 2.37(0.0204) ⇒ 0.8104 ± 0.0483 ⇒ (0.7621, 0.8587).
13.65 n = 17, ∑ x = 821, ∑ x2 = 43447, ∑ y = 873, ∑ y2 = 46273, ∑ xy = 40465,
⎡ (821)(873) ⎤
Sxy = 40465 − ⎢ ⎥ = 40465 − 42160.7647 = −1695.7647
⎣ 17 ⎦
⎡ (821)2 ⎤
Sxx = 43447 − ⎢ ⎥ = 43447 − 39649.4706 = 3797.5294
⎢⎣ 17 ⎥⎦
⎡ (873)2 ⎤
Syy = 46273 − ⎢ ⎥ = 46273 − 44831.1176 = 1441.8824
⎣⎢ 17 ⎦⎥
−1695.7647
b= = −0.4465
3797.5294
a = 51.3529 − (−0.4465)(48.2941) = 72.9162
SSResid = 46273 − 72.9162(873) − (−0.4465)(40465) = 684.78
388
684.78
se2 = = 45.652 , se = 6.7566
15
6.7566
sb = = 0.1096
3797.5294
a. Let β denote the average change in percentage area associated with a one year increase in
age.
Ho: β = −0.5 Ha: β =/ −0.5
α = 0.10
b− ( −0.5)
t= with df. = 15
sb
−0.4465 − ( −0.5)
t= = 0.49
0.1096
Since the P-value exceeds α, H0 is not rejected. There is not sufficient evidence in the
sample to contradict the prior belief of the researchers.
b. When x* = 50, y = 72.9162 + (−0.4465)(50) = 50.591
1 (50 − 48.2941)2
s a+b(50) = 6.7471 + = 1.647
17 3797.5294
The 95% confidence interval for the true average percent area covered by pores for all 50
year-olds is
50.591 ± (2.13)(1.647) ⇒ 50.591 ± 3.508 ⇒ (47.083, 54.099).
13.66 a. The summary values are: n = 10, ∑ x = 25, ∑ x2 = 145, ∑ y = −0.4, ∑ y2 = 43.88,
∑ xy = 55.5.
⎡ (25)( −.4) ⎤
55.5 − ⎢ ⎥ 56.5
b= ⎣ 10 ⎦= = 0.68485
⎡ (25)2 ⎤ 82.5
145 − ⎢ ⎥
⎣ 10 ⎦
a = −0.04 − 0.68485(2.5) = −1.7521
The equation of the estimated regression line is y = −1.7521 + 0.68485x.
b. SSResid = 43.88 − (−1.7521)(−0.4) − (0.68485)(55.5) = 5.169985
5.169985
se2 = = 0.646248
8
389
⎡ 1 x2 ⎤ ⎡ 1 (2.5)2 ⎤
sa2 = sa2+ b(0) = se2 ⎢ + ⎥ = 0.646248 ⎢ + ⎥ = 0.113583
⎣⎢ n Sxx ⎦⎥ ⎣⎢10 82.5 ⎦⎥
sa = 0.113583 = 0.337
Ho: α = 0 Ha: α =/ 0
α = 0.05
a
The test statistic is: t = with df. = 8
sa
−1.7521
t= = −5.20
0.337
P-value = 2(area under the 8 df t curve to the left of −5.20) < 2(0.002) = 0.004.
Since the P-value is less than α, the null hypothesis is rejected. The data suggests that the y
intercept of the true regression line differs from zero.
c. The 95% confidence interval for α is
−1.7521 ± (2.31)(0.337) ⇒ −1.7521 ± 0.7785 ⇒ (−2.5306, −0.9736).
Since the interval does not contain the value zero, zero is not one of the plausible values for
α.
13.67 Summary values for Leptodactylus ocellatus: n = 9, ∑ x = 64.2, ∑ x2 = 500.78, ∑ y = 19.6,

∑ y2 = 47.28, ∑ xy = 153.36
From these: b = 0.31636, SSResid = 0.3099, ∑ ( x − x ) 2 = 42.82
Summary values for Bufa marinus: n = 8, ∑ x = 55.9, ∑ x2 = 425.15, ∑ y = 21.6, ∑ y2 = 62.92,

∑ xy = 163.63
From these: b = 0.35978, SSResid = 0.1279, ∑ ( x − x ) 2 = 34.549
2 0.3099 + 0.1279 0.4378

s = = = 0.0337
9+8−4 13
H o : β = β′ H a : β ≠ β′
α = 0.05
b − b′
The test statistic is: t = with df. = 9 + 8 − 4 = 13.
2 2
s + s
SS x SSx′
0.31636 − 0.35978 −0.04342

t= = = −1.03
0.0337 0.0337 0.04198
+
42.82 34.549
390
P-value = 2(area under the 13 df t curve to the left of −1.03) = 2(0.161) = 0.322.
Since the P-value exceeds α, the null hypothesis of equal regression slopes cannot be rejected. The
data are consistent with the hypothesis that the slopes of the true regression lines for the two different
frog populations are identical.
13.68 For data set 1:
The plot above supports the appropriateness of fitting a simple linear regression model to data
set 1.
For data set 2:
This plot suggests quite clearly that the fitting of a simple linear regression model to data set 2 would
not be appropriate.
391
For data set 3:
This plot reveals an observation which would have an unusually large residual. A simple linear
regression model would not be appropriate for this data set, but might be for the data set with the one
unusual observation deleted.
For data set 4:
This plot reveals one point that is a very influential point. In fact, the slope is determined solely by
this observation. The simple linear regression model would not be appropriate for this data set.
13.69 When the point is included in the computations, the slope will be negative and much more extreme
(farther from 0) than if the point is excluded from the computations. Changing the slope will also have
an effect on the intercept.
13.70 Summary values are: n = 8, Sxx = 42, Syy = 586.875, Sxy = 1.5
1.5
a. b= = 0.0357 a = 58.125 − (0.0357)(4.5) = 57.964
42
The equation of the estimated regression line is y = 57.964 + 0.0357x.
392
b. Let β denote the expected change in glucose concentration associated with a one day
increase in fermentation time.
Ho: β = 0 Ha: β =/ 0
α = 0.10
b
The test statistic: t = with df. = 6.
sb
0.0357
From the data, sb = 1.526 and t = = 0.023.
1.526
P-value = 2(area under the 6 df t curve to the right of 0.023) = 0.982.
Since the P-value exceeds α, the null hypothesis is not rejected. The data does not indicate
a linear relationship between fermentation time and glucose concentration.
c.
x y Pred-y
Residual
1 74 58.00 16.00
2 54 58.04 −4.04
3 52 58.07 −6.07
4 51 58.11 −7.11
5 52 58.14 −6.14
6 53 58.18 −5.18
7 58 58.21 −0.21
8 71 58.25 12.75
d. The residual plot has a very distinct curvilinear pattern which indicates that a simple linear
regression model is not appropriate for describing the relationship between y and x. Instead,
a model incorporating curvature should be fit.
13.71 Since the P-value of 0.0076 is smaller than most reasonable levels of significance, the conclusion of
the model utility test would be that the percentage raise does appear to be linearly related to
productivity. This should be confirmed by examining a scatter plot of percentage raise versus
productivity.
393
13.72 a Summary values are: n = 17, Sxx = 13259.0706, Syy = 1766.4706, Sxy = −79.9294,
where x = depth and y = zinc concentration.
−79.9294
r= = −0.0165
(13259.0706)(1766.4706)
Let ρ denote the correlation between depth and zinc concentration.
Ho: ρ = 0 Ha: ρ =/ 0
α = 0.05
r
t= with df = 15.
1− r 2
(n− 2)
( −0.0165)
t= = −0.06
1 − ( −0.0165)2
15
P-value = 2(area under the 15 df t curve to the left of −0.06) ≈ 2(0.47) = 0.94.
Since the P-value exceeds α, the null hypothesis is not rejected. The data suggests that no
correlation exists between depth and zinc concentration.
b. Summary values are: n = 17, ∑ x = 531.7, ∑ x2 = 29,888.77, ∑ y = 58.5,

∑ y2 = 204.51, ∑ xy = 1983.64, Sxx = 13259.0706, Syy = 3.2012, Sxy = 153.9665,
where x = depth and y = iron concentration.
153.9665
r= =0.747
(13259.30706)(3.2012)
Let ρ denote the correlation between depth and iron concentration.
Ho: ρ = 0 Ha: ρ =/ 0
α = 0.05
r
t= with df. = 15
1− r 2
(n− 2)
0.747
t= = 4.35
1 − (0.747)2
15
Since the P-value is less than α, the null hypothesis is rejected. The data does suggest a
correlation between depth and iron concentration.
394
153.9665
c. b= = 0.01161
13259.0706
58.5 − (0.01161)(531.7)
a= = 3.0781
17
d. When x* = 50, a+b(50) = 3.6586.
SSResid = 204.51 − (3.0781)(58.5) − (0.01161)(1983.64) = 1.411
2 1.411
se = = 0.0941
15
⎡ 1 (50 − 31.276)2 ⎤
sa2+ b(50) = 0.0941⎢ + ⎥ = 0.00802
⎣⎢17 13259.0706 ⎦⎥
The 95% prediction interval for the iron concentration of a single core sample taken at a
depth of 50m. is
3.6586 ± (2.13) 0.0941 + 0 .00802 ⇒ 3.6586 ± (2.13)(0.3196)

⇒ 3.6585 ± 0.6807 ⇒ (2.9778, 4.3392).
e. When x* = 70, a+b(70) = 3.8908
⎡ 1 (70 − 31.276)2 ⎤
sa2+ b(70) = 0.0941⎢ + ⎥ = 0.0162
⎣⎢17 13259.0706 ⎦⎥
s a+b(70) = 0.0162 = 0.1273
The 95% confidence interval for α + β(70) is
3.8908 ± (2.13)(0.1273) ⇒ 3.8908 ± 0.2711 ⇒ (3.6197, 4.1619).
With 95% confidence it is estimated that the mean iron concentration at a depth of 70m is
between 3.6197 and 4.1619.
13.73 Even though the P-value is small and r2 = 0.65, the variability about the least-squares line appears to
increase as soil depth increases. If this is so, a residual plot would be “funnel shape”, opening up
toward the right. One of the conditions for valid application of the procedures described in this
chapter is that the variability of the y’s be constant. It appears that this requirement may not be true
in this instance.
13.74 a. The ei's are the deviations of the observations from the population regression line, whereas
the residuals are the deviations of the observations from the estimated regression line.
b. The simple linear regression model states that y = α + βx + e. Without the random deviation
e, the equation implies a deterministic model, whereas the simple linear regression model is
probabilistic.
c. The quantity b is a statistic. Its value is known once the sample has been collected, and
different samples result in different b values. Therefore, it does not make sense to test
hypotheses about b. Only hypotheses about a population characteristic can be tested.
395
d. If r = +1 or −1, then each point falls exactly on the regression line and SSResid would equal
zero. A true statement is that SSResid is always greater than or equal to zero.
e. The sum of the residuals must equal zero. Thus, if they are not all exactly zero, at least one
must be positive and at least one must be negative. They cannot all be positive. Since there
are some positive and no negative values among the reported residuals, the student must
have made an error.
f. SSTo = ∑ (y − y )2
must be greater than or equal to SSResid = ∑ (y- ŷ )2. Thus, the values
given must be incorrect.
13.75 a. When r = 0, then se ≈ sy. The least squares line in this case is a horizontal line with intercept
of y .
b. When r is close to 1 in absolute value, then se will be much smaller than sy.
396

Chapter 1 Exercises on Descriptive Statistics and Data Types

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Chapter 1 Exercises on Descriptive Statistics and Data Types

Uploaded by

Copyright:

Available Formats

Chapter 1

Exercise 1.1 – 1.9

Exercise 1.10 – 1.26

1.10 a. numerical (discrete)

Breakdown of all Complaints, by Type

1.18 a. Categorical. b. No, because a dotplot is used for numerical data.

Literacy Level in Adults as Classified by the National

"Plagiarism is a significant Problem in Academic

Causes of Children's Weight Problems

Reasons for Leaving Before Graduating

Sleepy at Work? Relative Frequency

Make-up of American Suburbs

Sport Frequency Rel.

Reasons for Nonresponse

2.1 a. It is an observational study, since no treatment was imposed.

b. No. Cause-and-effect conclusions cannot be made from an observational study.

Exercises 2.10 – 2.30

2.15 1. Cluster sampling

e.& f. answers will vary

Exercises 2.31 – 2.39

2.32 a. Strength of binding.

2.35 a. Blocking. Homogeneous groups were formed by height.

b. Direct control. The researchers only used right handed people.

iv. It is impossible to blind by anything obvious – gender, age, race.

b. To make the two groups as similar as possible to reduce any bias

b. So that the evaluation is not swayed by personal interest.

c. As irritability is so subjective, double-blinding, as described in a, would be advisable.

Exercises 2.51 – 2.56 Answers will vary

Exercises 2.57 – 2.70

2.57 a. It is an observational study; no treatment was imposed.

2.58 a. It is an observational study; no treatment was imposed.

b. No. No cause-and-effect can be concluded from an observational study.

2.60 a. Yes, because of the random assignment of children to treatment.

b. No; children were not randomly selected.

Change in Tobacco Smoking 1999 to 2001

What can you survive without?

Percentage of Students passing the Exit Exam by district

Christmas shopping from Mail Order Catalogues

Christmas online shopping

Roadside Hazard by Object Struck

Exercises 3.22 to 3.34

Class Interval Frequency Relative Frequency

Histogram of Cost per month (in dollars)

Credit Card Balance (in $) Credit Card Bureau Data)

b. .4167 + .1792 + .1500 = .7459

d. .1000 + .0375 + .0458 = .1833

Average # Cell phone mins used per Month by Men

0 200 400 600 800

Average # Cell phone mins used per Month by Women

0 200 400 600 800

i 0.93 ii 0.34 iii 17 mins

stem: thousands digit

Classes Frequency Rel. Freq. Cum. Rel. Freq.

ii. Approximately 29 months

This histogram is symmetric.

This histogram is positively skewed.

This is a bimodal histogram.

This is a negatively skewed histogram.

a. The histogram associated with frequency set I would be symmetric.

b. The histogram associated with frequency set IV would be bimodal.

d. The histogram associated with frequency set II would be negatively skewed.

Exercises 3.35 to 3.45