5.5 Hetroskedasticity Test
Heteroskedasticity test is also done for the model I and the results look like seen below in Table 5.12. Since the Obs*R2 value of 8.092 is less than the 5% critical X2 value of 11.07, the null hypothesis that assumes unavailability of heteroskedasticity can’t be rejected. That implies that the standard errors, T-statistics and F-statistics can be considered valid.
Table 5. 12 White Heteroskedasticity Test Result for Model I
F-statistic 1.410 Probability 0.250
Obs*R-squared 8.092 Probability 0.231
The result of heteroskedasticity test done for Model II is also shown in Table 5.13 below. The null hypothesis of no heteroskedasticity cannot be rejected this time too. This result is the same as the preveous model’s result and the post-regression test results are valid.
Table 5. 13 Heteroskedasticity Test Result for Model II
F-statistic 0.714 Probability 0.722
Obs*R-squared 9.893 Probability 0.625
5.6 T test for Coefficient Significance
The government expenditure variables were hypothesized to have a positive effect on GDP of the country. This implies that the coefficients on those independent variables are expected to be positive and a one-tailed test is appropriate.
For Model I, the T-test results (Table 5.14) showed that log(RE) and log(CE), which represent recurrent and capital expenditure respectively are relevant variables while foreign aid is not.
Table 5. 14 Result of T test for
We first started by examining the different variables in relation to rounds, and seeing their corresponding p-values. Since YARD, RANGE, and WINTER*FEE all have insignificant p-values, we remove them from our model in order to increase the accuracy, Using the Wald test, we confirm they are irrelevant to the model, and are taken out. Then, using our new model, we must test for heteroscedasticity using the White test. We find there is, and so we adjust our model accordingly. Finally, we reanalysed all the variables and made sure they were all still logically correct, and conclude our model is BLUE. From here, we computed
4. Based on your analysis in (1) – (3) above, what is your overall conclusion regarding the
b. Then, we also determine the regression equation and correlation between North's engine cost and the average age of fleet. Please refer to Annex 3 for the QM results.
For d2, t-statistic= 1.8774, t-statistic < t-critical. Thus we do not reject Ho and d2 is not significant.
Referring to Figure 6.2.3 and 6.3.3, it was proven that our model has problems that the sample data used does not represent the whole population. Therefore, this is one of the flaws in our research. A more constructive suggestion to eliminate this problem would be to extend the research with a larger sample size with longer time horizon. And if the sample size is large enough, the time series issue can be neglected.
My original model was able to pass most of the assumptions but not all. The error terms were normal (Figure 1) and there were no serious outliers, multicollinearity, or autocorrelation. However, the scatter plot of the predicted Y values and the residuals showed signs of heteroscedasticity, depicted in Figure 2. Given that, I transformed my dependent variable by squaring the values of the dependent variables. This corrected the heteroscedasticity, as shown in Figure 3 and the error terms still followed a normal distribution, as given by Figure 4. My final model now checked off all of the assumptions and I could move forward.
Model Fit Summary CMIN Model | NPAR | CMIN | DF | P | CMIN/DF | Saturated model | 36 | .000 | 0 | | | Independence model | 8 | 3797.971 | 28 | .000 | 135.642 | RMR, GFI Model | RMR | GFI | AGFI | PGFI | Saturated model | .000 | 1.000 | | |
10. Refer to the model and estimates in the previous question. You want to test that
a. What are the estimated sales for the Bryne store, which has four competitors, a
To avoid spurious results, unit root tests using Augmented Dickey-Fuller (ADF) (Dickey and Fuller, 1981) and Philips-Perron (PP) are performed to determine the time-series properties of the variables employed in the analysis. Two or more variables are said to be co-integrated when they exhibit long run equilibrium (relationship) if they share common trend(s).Therefore Auto-regressive distributed lag bounds approach (ARDL) is used to test it. The choice is based on several
You provided no discussion about TAXYPR, TGEG, and DMC in Table 4. How can you insert TGEG and DMC into the regression equation as explanatory variables?
B) A test statistic of t = 1.813 with d.f. = 15 leads to a clear-cut decision.
* Test the utility of this regression model (use a two tail test with α =.05).
As mentioned in class, one commonly employed solution to heteroscedasticity is to adjust the standard errors for the possible presence of heteroskedasticity, i.e. we compute the heteroskedasticity-robust standard errors, which are also referred to as heteroskedasticity-consistent standard errors. Rerun the regression in part (2) with the OLS standard errors replaced by heteroskedasticity-robust standard errors. Comment on the differences between the OLS standard errors in part (2) and the heteroskedasticity-robust standard errors in this part.
Table 4.1 presents the panel unit-root test results. There are two groups of hypotheses that are involved here. In the first four methods, the null hypothesis is: there is panel unit-root and the alternative hypothesis is: there is no panel unit-root and the decision