1.0 0.8
0.6 0.4
0.2 0.0
Observed Cum Prob
1.0 0.8
0.6 0.4
0.2 0.0
E xpect
ed C
um P
rob Dependent Variable: Y
Normal P-P Plot of Regression Standardized Residual
1.0 0.8
0.6 0.4
0.2 0.0
Observed Cum Prob
1.0 0.8
0.6 0.4
0.2 0.0
E xpect
ed C
um P
rob
Dependent Variable: Y Normal P-P Plot of Regression Standardized Residual
- Indicator 8
- Indicator 9
- Indicator 10
- Indicator 11
0,920 0,920
0,760 0,862
0,198 0,198
0,198 0,198
Valid Valid
Valid Valid
Source: primary data, 2010
From tables 4.4 obtained that all indicators used to measure the applied variables in this research has higher correlation coefficient than r table = 0,198 r table value of n=100.
So, all indicators are valid.
4.5. Classic Assumption Test
This research applies two multiple linear regression models. A good regression model must free of classic assumption problems. Following is classic assumption test for both
regression models.
4.5.1. Normality Test
Normality test was done by using testing to residual value. While the test was done by using P-P Plot. Normality test result can be visibly seen from the following figure.
Figure 4.5 Normality Test Result
Model 1 Model 2
.564 1.773
.579 1.726
.582 1.718
.447 2.239
.578 1.730
.591 1.692
.469 2.132
.390 2.562
.328 3.045
X1 X2
X3 X4
X5 X1X5
X2X5 X3X5
X3X5 Model
1 Tolerance
VIF Collinearity Statistic s
.615 1.626
.623 1.606
.613 1.631
.517 1.933
X1 X2
X3 X4
Model 1
Tolerance VIF
Collinearity Statistic s
Source: primary data, 2010
The figure indicates that residual points from both regression model have normal distribution because the points disseminating around diagonal line. Thereby normality
condition required as statistical testing by using regression can be fulfilled.
4.5.2. Multicollinearity Test
A variable which shows multicollinearity symptoms can be seen from its high VIF Variance Inflation Factor value in a regression model. VIF value higher than 10 showing
the existence of multicollinearity symptom in modeling regression. Result of VIF test from
both regression models are as follows: Table 4.22
Multicollinearity Test Model 1
Model 2
Source: primary data, 2010
Result of the test indicates that all variables applied as regression model predictor shows sufficiently small VIF values, where altogether below number 10. It means that free
variables applied in research doesnt show existence of multicollinearity symptom, then each independent variables serve as independent predictor.
4.5.3. Heteroskedasticity Test
Heteroskedasticity test was done using scatter plot. If there is no regular pattern at its residual points, hence no heteroskedasticity problem detected. Result of the test shown in the
following figure.
Figure 4.6 Heteroskedasticity Test Result
Model 1
3 2
1 -1
-2 -3
Regression Standardized Predicted Value
4 3
2 1
-1 -2
-3
R egressi
on S
tudent iz
ed R
esi dual
Dependent Variable: Y Scatterplot
Model 2
Source: primary data, 2010
Heteroskedasticity test result shows there is no independent variable which significantly relates to absolute residual value. It means that both regression models dont
have the symptom of heteroskedasticity existence.
4.6. Regression Analysis 4.6.1. Linear Regression Analysis Model 1 Testing of Hypothesis 1 - 4