Using Some Estimation Methods for Mixed-Random Panel Data Regression Models with Serially Correlated Errors with Application

Abstract

This research includes the study of dual data models with mixed random parameters, which contain two types of parameters, the first is random and the other is fixed. For the random parameter, it is obtained as a result of differences in the marginal tendencies of the cross sections, and for the fixed parameter, it is obtained as a result of differences in fixed limits, and random errors for each section. Accidental bearing the characteristic of heterogeneity of variance in addition to the presence of serial correlation of the first degree, and the main objective in this research is the use of efficient methods commensurate with the paired data in the case of small samples, and to achieve this goal, the feasible general least squares method (FGLS) and the mean group method (MG) were used, and then the efficiency of the extracted estimators was compared in the case of mixed random parameters and the method that gives us the efficient estimator was chosen. Real data was applied that included the per capita consumption of electric energy (Y) for five countries, which represents the number of cross-sections (N = 5) over nine years (T = 9), so the number of observations is (n = 45) observations, and the explanatory variables are the consumer price index (X1) and the per capita GDP (X2). To evaluate the performance of the estimators of the (FGLS) method and the (MG) method on the general model, the mean absolute percentage error (MAPE) scale was used to compare the efficiency of the estimators. The results showed that the mean group estimation (MG) method is the best method for parameter estimation than the (FGLS) method. Also, the (MG) appeared to be the best and best method for estimating sub-parameters for each cross-section (country).