The Comparison of Familiar Estimators in Panel Model

The comparison of familiar estimators of regress coefficients in panel model be considered in this paper. Under the hypothesis that error vector are distributed as multivariate normal, some significative results can be obtained by using the excellent property of multivariate normal. By Pitman criterion, the sufficient condition under which least square estimator is better than within estimator be obtained, and then least square estimator is uniformly better than within estimator. By generalized mean squared errors criterion, we obtain a useful result that Two-Stage estimator is better than between estimator. Finally, we give two direct corollaries about Two-Stage estimator is better than between estimator and within estimator.