Panel模型中常见估计的比较  

The Comparison of Familiar Estimators in Panel Model

作　　者：马铁丰[1] 王松桂[1] 

机构地区：[1]北京工业大学应用数理学院,北京100022

出　　处：《数学进展》2008年第1期107-114,共8页Advances in Mathematics（China）

基　　金：北京市属市管高等学校人才强教计划资助项目(No.05006011200702).

摘　　要：本文研究了Panel模型中回归系数常见估计的比较问题,给出了在Pitman准则,协方差阵准则和广义均方误差准则下最小二乘估计,Within估计,Between估计及两步估计之间的优良性比较结果.特别地,本文证明了在Pitman准则下最小二乘估计一致地优于Between估计.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.

关 键 词：Panel模型 PITMAN准则 协方差阵 广义均方误差 两步估计 

分 类 号：O212.1[理学—概率论与数理统计]