Following are the assumptions of multiple regression analysis:
Independence of observation Independence of observation assumption implies that there should not exist any relationship between the responses of any respondent with that of another respondent (Leech, Barret & Morgan, 2005). Durbin-Watson test is used to test the check the presence of any serial correlation among the residuals of the observations. Durbin-Watson test statistics for the regression test of this study is 2.11. The values range from 0 to 4 however optimally it should be between 1.5 and 2.5. Normal Distribution of Data The assumption of normal distribution of data implies that when the data is plotted; it is symmetrically dispersed around the mean value of the data. The data for this study was normally distributed as shown in the histogram graphs for normal distribution. The graphs show a normal distribution in the bell curve for skewness and kurtosis. Morgan et al. (2001) stated that the skewness and kurtosis statistics can be interpreted as symmetrical as long as they are between the values of +1 and -1.
Figure 2. Histogram chart for the Personality variable.
Hair et al. (2006) stated that the Q-Q plot also determines the normality of data. If the data plotted is along the normal line then the data is considered as normal.
Figure 6. Normal Q-Q plot
Outliers The assumption of outliers suggests that there should not be a unique or distinct response from the rest of the data. It usually exists when there is one value on either of the extreme sides of the x-axis and deviates from the other values of the sample (F ield, 2005). This can be detected by the z-score using SPSS (Brace, Kemp & Snelgar, 2006) and it should not go beyond the limit of +2.5 and -2.5 (p<0.05, two tailed). Linearity of Data The linearity of data assumes that there exists a linear relationship between all the dependent and independent variables. It means that when the variables are plotted in a scatter plot diagram, there should be a linear relationship between them (Hair, et al., 2006; Tabachnick & Fidell, 2001).
Figure 10. Scatter plot diagram for the variables
In addition, the scatter plot matrix can also be used to visually represent the relationship of variables that is shown in the figure 11.
Figure 11. Scatterplot Matrix for the Variables
Homoscedasticity Homoscedasticity is assumes that the variance of all the dependent variables is somewhat equal with all the independent variables. The scatter plot of the residuals of the regression
analysis also shows the homogeneity of variances as long as they are equally spread along the zero line.
Figure 12. Scatter plot graph of the Regression Residuals
Mulitcollinearity and Singularity The assumption of multicollinearity and singularity tends to be violated, if there is found a very high correlation among the independent variables in the correlation matrix. If the correlation is very high (i.e. 0.90 or above) there exists the problem of multicollinearity however, the problem of singularity is found if there is a perfect correlation among different variables of the study. In this study, there is not found any problem of mulitcollinearity or singularity as all the correlation values are less than 0.90 (see table 7). However, multicollinearity can be checked by the tolerance value and variance inflation factor (VIF) obtained from collinearity matrix. Field (2005) discussed that the tolerance value should be greater than 0.20 and VIF value sho uld be less than 4 in order to avoid the problems of multicollinearity and singularity. The tolerance values and VIF values for the variables of this study are shown in table 20. It shows that the values of the collinearity statistics suffice the assumptio n of multicolinearity and singularity.