You are on page 1of 16

DATA ANALYSIS AND

INTERPRETATION
• The data submitted for analysis and interpretation. In particular, we
will examine the following:
• 1. The establishment of Cronbach‘s alpha for the measures.
• 2. The frequency distribution of the variables.
• 3. Descriptive statistics such as the mean and standard deviation.
• 4. The Pearson correlation matrix.
• 5. The results of hypotheses testing.
• Checking the Reliability of Measures: Cronbach‟s Alpha
• Obtaining Descriptive Statistics: Frequency Distributions
• Descriptive Statistics: Measures of Central Tendencies and
Dispersion
Unit of Analysis
Based on the objectives and questions developed in the study, it is found
appropriate to select the unit of analysis in the study.

Degree of Involvement
The respondent’s consent will be obtained for data collection using survey and
interview approach. This technique will be adopted to avoid interruption in the
routine task of the respondents.
Moreover, analysis will be done without involving the respondents to minimize
the degree of involvement of the respondent. This will avert the impact of
unnecessary manipulation on the study.
Inferential Statistics:
Data Screening
• Data screening is important because it make sure that the data
collected during the field survey is good enough to pass through
various analyses and tests needed to answer the hypothesis
developed for the study. It would have no importance if it proposes
vague solutions (Davies, 2005; s et al., 2006). Data screening will pass
through the following stages:
1. Missing Value Analysis
• Sekaran (2003) proposed to use imputation procedure to substitute the cells
which will be left empty. According to this method the mean value will be
replaced with the missing value found in some specific cell. The idea has also
got support from Meyers et al. (2006) as they considered it to be the most
conventional way of replacing the missing values.
2. Extreme Value or Outliers Examination
• This test will be used to refute the impact of extreme value on overall mean
of the item (Sekaran, 2003). Outlier examination will be used to enhance the
quality and clarity of out puts obtained through factor analysis (Meyers et
al., 2006). Different methods are used to recognize the extreme values for
example Z-score and Box-plot. According to Hair et al (2006) Z-score is a
more convincing test. Outliers are thought to be exists if the Z-value
exceeds + 4.0σ.

Multivariate Test
• In order to follow the set standards, different multivariate analyses processes
are required to be completed to make the data ready for regression analysis.
This will be done to confirm that the findings obtained are able to explain the
phenomenon significantly (Meyers et al., 2006). Multivariate tests used for this
research are discussed as follows.
Normality Test
• There are different methods used to identify the normality of
the distribution. Meyers et al. (2006) proposed that skewness
and kurtosis can be used as indicator to check the deviation.
Data is said to be normally distributed if the range of
skewness and kurtosis lies within + 1.0. Field (2005) is of the
opinion that Kolmogorov-Smirnov test (K-S test) is one of a
good indicator to determine the deviation of the distribution
from normality. A distribution is considered to be normal if
the KS-test result shows that the sample is not significantly
different from a normal distribution (i.e. p > 0.05).
Multicollinearity Test
• To measure the strength of correlation, in the regression model,
between two or more variables, a multicollinearity test will be
applied.
• Strong collinearity makes it difficult to get precise estimation. It also
escalates the b-coefficient’s standard errors in the model and hence
threatens the model; thus will affect the statistical impact of
coefficients (Field, 2005).
• Therefore, the collinearity’s strength may enhance the chances of
type II error because due to strong collinearity, there is a chance of
rejection of a good hypothesis.
• Multicollinearity among the variable (the independent variables) will
be calculated using Tolerance and Variance Inflation Factor (VIF).
• A tolerance value nearer to 1 will show that multicollinearity among the
variables is less, on the other hand the value near to 0 shows that
multicollinearity can be a problem. According to Meyers et al. (2006) there
exists multicollinearity among the variable if the value of tolerance becomes
less than 0.01.

• Variance Inflation factor (VIF), which is the reciprocal of the tolerance, shows
the impact of multicollinearity on variance of the coefficient estimate. In other
words, the value of VIF close to Zero shows that there is no correlation
between the variables.
• Hence, smaller value for VIF is desirable to detect the multicollinearity.
• Scholars are, however, indecisive regarding the desired value of VIF as an
indicator of multicollinearity among the independent variables (IVs). According
to Hair et al. (2006) VIF value of 5.0 could be considered as a value indicating
the non-existence of multicollinearity. Whereas Meyers et al. (2006) suggests
this value as 10.0.
Correlation Analysis
Correlation coefficient with a value of + 0.5 is expected to have a greater effect.
Multiple Regression Analysis
• In order to observe the influence of different variables (independent variables) on the other
variables (dependent variables) researchers used Multiple linear regression analysis
technique in which impact of multiple factors can be measured simultaneously (Field, 2005).
• While applying this technique, three assumptions needed to be fulfilled which are,
independence of errors, normality of error and equality of variance (Field, 2005).
• The first assumption entails that the errors should be independent of each other. This
assumption is important in the situation where data collection is spread over a time;
however, the normality assumption could not be neglected. Further, there should not be any
correlation between independent variables (Field, 2005).
• The tests for normality of error, multicollinearity and hetroscedasticity will be used to ensure
that all required assumptions are fulfilled. On the basis of application of multiple regression
analysis, following will be the regression model:
DV= b (IV) + Constant
• In the Regression Analysis, the Model Summary gives the values of R, R square, adjusted R square,
and the standard error.

• R gives the correlation effect of observed and predicted values of the dependent variable.
The R-values range is from -1 to +1. The positive value of R specifies that there exists a
positive relationship between the variables. It is a generally accepted principle that the
larger the R value stronger will be the relationships.
• The value of R square explains the proportion of variance in the dependent variable
described by the regression model. The R-square value range is 0 to 1. If the model gives
small value for R-square, it shows that the model is weak. In other words, it estimates “how
well the model fits the population”.
• The adjusted R square values try to make corrections in the R square values to make the
model a better fit while closely reflecting the goodness of fit of the model in the population.
Usually the values of R Square are used to compare the models and to determine which
model is best. Therefore, the model with higher R-square value is chosen a good fitted
model.
While conducting multiple regression analysis, it is important to “test whether the model
generated is significantly better at predicting the outcome than using the mean as a ‘best
guess’ method” (Field, 2005, p. 189). This information will be discussed using ANOVA
table in the regression analysis where a significant value pointed out that the model is a
significant fit and is not appeared by chance.

The Analysis of Variance (ANOVA) explains whether the independent variable performed
a good job in explaining the variations in the dependent variable or not. For this purpose
the significance value of F should have a value smaller than 0.05.

Table ANOVAb
Model Sum of Squares df Mean Square F Sig.
1 Regression 5.222 1 5.222 40.200 .000a
Residual 18.317 141 .130
Total 23.540 142
a. Predictors: (Constant), Industrial Growth
b. Dependent Variable: Economic Competitiveness
Mediation Test
Baron and Kenny (1986) proposed that mediation test is completed in four steps.
In the first step,
the dependent variable (DV) should significantly be predicted by the independent variable (IV).
In the second step,
mediating variable should significantly be predicted by the IV and
In the third step
the prediction of DV should be significant using mediator.
In the fourth and final step,
the DV should also be predicted significantly by both IV and mediator collectively.

If this four step criteria is met, then the independent variable’s direct effect definitely be
reduced. If this effect of reduction of IV reached to zero, complete mediation has been
confirmed. In this situation the significance of the mediation model will be tested using a Sobel
test (MacKinnon and Dwyer, 1993; Soper, 2011).
Sobel Test
• In order to measure the mediating effect, Sobel test for mediation will be
performed. To perform the Sobel test, ‘raw regression coefficient’ and
‘standard error’ for this coefficient was computed to measure the
relationship between Independent Variable and the Mediator, and the
relationship between the Mediator and the Dependent Variable (adjusting
for the IV).
• The values are than entered in online table at their respective places; the
output appears as given in Figure
Un-Stand. Stand. 2 Sobel’s
Steps Model Name R Sig.
B S. E Beta T-Stat P-Val
Model 1. Industrial Growth (IV)-University (MV)-Economic Competitiveness (DV)
Step 1 IV-DV 0.381 0.060 0.471 .222 0.000
(Constant) 2.365 0.314
Step 2
IV-MV 0.403 0.082 0.383 .147 0.000
Step 3 MV-DV 0.467 0.051 0.608 .370 0.000
IV-DV 0.225 0.056 0.279 0.000
Step 4
MV-DV 0.385 0.053 0.501 0.000
Sobel Test 5.46 0.000
Model 2. Prof. Development (IV)-University (MV)-Econ Competitiveness (DV)
Step 1 IV-DV 0.358 0.049 0.521 .272 0.000
(Constant) 2.350 0.257
Step 2
IV-MV 0.409 0.067 0.456 .208 0.000

You might also like