4.3. Factor Analysis In order to meet the objective of the research, EFA has been used to reduce the number of variables to a smaller set of factors (Hair et al., 2006; Field, 2013) and to minimize the problem of multicollinearity (Malhotra and Dash, 2011). The motive is to find the degree of dependency of these individual variables on the dependent variable.
The whole set of data has been compiled in SPSS 21. The execution of the project started with the grouping of the variables using factor analysis, and then the relationship between the dependent variable and the predictors has been established using multiple regression analysis. The output in the form of regression equation has been analyzed at the last of thetopic.
The variables identified in the research and the data sample collected on these variables are input to EFA. Through EFA, the number of variables is reduced to a fewer set of factors using varimax rotation. The EFA technique is widely used for the reduction of the problem of multicollinearity among the variables. The criteria for EFA for finalizing the factor structure are as follows (Hair et al., 2006): 1 Correlation matrix and Anti-Image Matrix results show how suitable the data is for the EFA technique. These have been used as a tool to filter out the variables not adequate for factor analysis (Field, 2013).
2 Kieser-Meyer-Olkin (KMO) value and Bartlett’s test for sphericity are used as measures of sampling adequacy. To meet the criteria, the KMO value must be more than 0.8 With the KMO value of 0.922, and Bartlett’s significance value of 0.
000, the sample collected is extremely adequate (Hair et al., 2006; Field, 2013). 3 Further the factor loading defines the amount of correlation between the variables and the factors to which they belong. The factor loading value can vary from –1 to +1 and square of any of these numbers defines the amount of variability accounted for this factor.
Lower factor loading values, i.e. less than 0.5 and lower communality variables are filtered from the data set (Hair et al., 2006; Field, 2013). 4 The Eigen value amounts the total variance explained by the factors.
It is also a kind of filtering technique, as the factors with Eigen values greater than 1 are considered. As we can see in the table below, there are in all 6 factors which have Eigen value greater than 1 (Hair et al., 2006). 5 The percentage variance defines the percentage of total variance being explained by that particular factor.