Common Effect or Fixed Effect Model Fixed Effect FEM or Random Effect Model REM

3.5.2.1. Common Effect or Fixed Effect Model

F-statistic test is done to examine whether in panel data regression, Fixed Effect is better than panel data regression model without dummy variable Common Effect by looking for the value of Sum of Squared Residual SSR. The equation of F-statistic test is as follows: F = where: SSR R = Sum of squared residuals of common effect model estimation result SSR U = Sum of squared residuals fixed effect model estimation result q = Number of common effect model CEM restriction n = Number of observations k = Number of parameters in the Fixed Effect Model FEM F-statistic values compared with the F-statistic critical degrees of freedom df for the numerator and q as many as n - k for denumerator. If the F-statistic is greater than the F-critical, then used a Fixed Effect Model. If the F-statistic smaller F-critical, then the Common Effect Model was used. Likelihood Ratio test is a test conducted by Eviews8 to simplify the selection between Common Effect Model and Fixed Effect Model. The hypothesis of this test is as follows: H : Fixed Effect model is equal to Common Effect. H 1 : Fixed Effect model is better than Common Effect. In other words, if the value of F is significant, it means that Fixed Effect model is better than Common Effect Ghozali, 2013.

3.5.2.2. Fixed Effect FEM or Random Effect Model REM

Hausman test is used to examine which model fit the best among Fixed and Random Effect. This test is done by comparing chi-square x 2 statistic with x 2 -critical value on the degree of freedom df as many as the amount of independent variables k Widarjono, 2013. Hypothesis used in Hausman test is: H : Random Effect Model H 1 : Fixed Effect Model If the value of chi-square x 2 -statistic is bigger than x 2 -critical, so the best model is Fixed Effect, and vice versa. The result estimation also can be seen from the significance of probability. If it shows significance probability, it means that Fixed Effect Model is chosen, and vice versa Ghozali, 2013.

3.5.2.3. Common Effect CEM or Random Effect Model REM