Normality Test Multiple Linear Regression Model

53 Explanation of Table 4.1: 1. The dependent variable of the stocks rate of return in consumer goods sector has an average return of 0.022268. The maximum return of consumer goods sector during the period of observation is 0.193360, while the minimum return is -0.169444 with the 6.785 standard deviation. In the meantime, the dependent variable of the stocks rate of return in property and real estate sector has an average return of 0.019215. The maximum return of property and real estate sector during the period of observation is 0.306532, while the minimum return is -0.340247with the 9.9661 standard deviation. 2. As for the independent variable, the change in exchange rate during the period of the observation has an average change of -0.000831. The maximum change is 0.172425 and the minimum change is -0.098840 with 3.7145 standard deviation. Another variable, which is change in inflation rate, has an average change of -0.483316. The maximum change is 9.000000 and the minimum change is -35.00000with 519.2304 standard deviation. And as the last independent variable, which is change in SBI rate, it has the average change of -0.011012. The maximum change is 0.130793 and the minimum change is -0.105425 with 3.6939 standard deviation.

C. Normality Test

In this research, the writer uses the normality test of Jarque Bera using statistical software Eviews 5. Normality test is used to test whether the residual of the model 54 used in the research is normally distributed or not. The diagram below present the result of normality test using histogram residualJarque-Bera method for the consumer goods sector: Figure 4.1 Histogram Residual, Consumer Goods Sector Figure 4.2 Histogram Residual, Property and Real Estate Sector 1 2 3 4 5 6 7 8 9 -0.10 -0.05 -0.00 0.05 0.10 0.15 Series: Residuals Sample 2006M01 2010M12 Observations 60 Mean -4.97e-18 Median -0.005805 Maximum 0.145997 Minimum -0.105790 Std. Dev. 0.057312 Skewness 0.487487 Kurtosis 2.684504 Jarque-Bera 2.625284 Probability 0.269108 2 4 6 8 10 12 14 -0.1 -0.0 0.1 0.2 Series: Residuals Sample 2006M01 2010M12 Observations 60 Mean 6.94e-19 Median 0.001880 Maximum 0.238803 Minimum -0.167392 Std. Dev. 0.079104 Skewness 0.363828 Kurtosis 3.382237 Jarque-Bera 1.688972 Probability 0.429778 55 From the Figure 4.1 we can see that the probability of Jarque Bera of the consumer goods sector is bigger than α 0.269108 0.05. Thus, we failed to reject the Null Hypothesis and that means the residual of the model is normally distributed. From the Figure 4.2 we can see that the probability of Jarque Bera of the property and real estate sector is bigger than α 0.429778 0.05. Thus, we failed to reject the Null Hypothesis and that means the residual of the model is normally distributed.

D. Classical Assumption Test

In order to achieve the model that has BLUE characteristic, the writer needs to conduct the classical assumption test before doing the regression analysis. If in this test we find that the model is not resulting in BLUE characteristic we will use the remedial measurement accordingly.

1. Heteroscedasticity Test

One of the important assumptions of the classical linear regression model is that the variance of each disturbance, conditional on the chosen values of the explanatory variables, is somewhat constant. This is the assumption of homoscedasticity. To detect whether heteroscedasticity is present in the data, the writer will conduct a formal test using White Test method. The reason why we use White‟s General Heteroscedasticity Test is because it does not rely on the normality assumption and easy to implement. 56 After using the statistical software Eviews 5, the writer find the result of White Heteroscedasticity Test- No cross terms, as follows: White Heteroskedasticity Test: F-statistic 1.985143 Probability 0.084204 ObsR-squared 11.00974 Probability 0.088076 Table 4.2 Heteroscedasticity Test for Consumer Goods Sector Table 4.2 presents the result of the White Heteroscedasticity test on consumer goods sector using statistical software Eviews 5. The result of the test provides the proof that the data we are using is free from heteroscedastic problem homoscedastic. From the test above, we can see that the probability of ObsR- squared is bigger than our significant value 0.088076 0.05 and thus we can reject the null hypothesis. White Heteroskedasticity Test: F-statistic 0.538939 Probability 0.776195 ObsR-squared 3.450216 Probability 0.750581 Table 4.3 Heteroscedasticity Test for Property and Real Estate Sector 57 Table 4.3 presents the result of the White Heteroscedasticity test on property and real estate sector using statistical software Eviews 5. The result of the test provides the proof that the data we are using is free from heteroscedastic problem homoscedastic. From the test above, we can see that the probability of ObsR- squared is bigger than our significant value 0750581 0.05 and thus we can reject the null hypothesis.

2. Auto-Correlation Test

As in the case of heteroscedasticity, in the presence of autocorrelation the OLS estimators are still linear unbiased as well as consistent and asymptotically normally distributed, but they are no longer efficient i.e., minimum variance. To detect autocorrelation, the writer use The Breusch-Godfrey BG Test, which is also known as Lagrange-Multiplier LM Test in the application software EViews 5. The results are presented in tables below: Breusch-Godfrey Serial Correlation LM Test: F-statistic 0.150862 Probability 0.860328 ObsR-squared 0.333385 Probability 0.846460 Table 4.4 Auto-Correlation Test for Consumer Goods Sector Breusch-Godfrey Serial Correlation LM Test: F-statistic 0.587759 Probability 0.559086 ObsR-squared 1.278303 Probability 0.527740 Table 4.5 Auto-Correlation Test for Property and Real Estate Sector 58 From the Table 4.4 we find that the probability ObsR-squared from the data of consumer goods sector is bigger than sig. value α 0.846460 0.05, thus we can conclude that auto-correlation is not existed in our data. The same result happens in the property and real estate sector. As if presented in Table 4.5, the probability ObsR-squared is bigger than sig. value α 0.527740 0.05, thus we can reject the null hypothesis and conclude that there is no auto-correlation in our data.

3. Multi-Collinearity Test

R-Squared Y R-Squared X 1 R-Squared X 2 R-Squared X 3 Consumer Goods 0.286679 0.170209 0.039652 0.142669 Property and Real Estate 0.369993 0.170209 0.039652 0.142669 Table 4.6 Multi-Collinearity Test When examining the existence of multi-collinearity, one should look at the partial correlation coefficient Gujarati, 2004. The table above describes the summary of the linear regression, given each variable to be the dependent variable. For the consumer goods sector, the R-Squared of the Dependent Variable Y is bigger than its independent variables y=28.6667 x 1 =17.0209 x 2 =3.9652 x 3 =14.26, and thus we reject H and assume that there is no multi-collinearity existed in the data. The same thing goes for Property and Real Estate Sector, where the obtained R-squared Y is 36.99, which is bigger than its independent variables. And thus, it is confirmed that multi- collinearity doesn‟t exists as in the property and real estate as well. 59

E. Multiple Linear Regression Model

Several classical assumption tests have done through the model to ensure that the model is a good estimator. In this research, the writer use the multiple linear regression model to estimate the influence of the inflation rate, exchange rate and SBI rate to the return of stocks in consumer goods and property and real estate sector. In order to achieve that, the writers use the statistical software Eviews 5 which is resulted in models as follow:

4.1 Model of Consumer Goods Sector

4.2 Model of Property and Real Estate Sector

F. Hypothesis Testing 1. T-Test