多元t分布下相依回归模型参数的两步估计  

TWO-STAGE ESTIMATORS OF SEEMINGLY UNRELATED REGRESSIONS WITH MULTIVARIATE t-DISTRIBUTION

作　　者：刘金山[1] 

机构地区：[1]华南农业大学理学院数学系,广州510642

出　　处：《系统科学与数学》2003年第2期198-204,共7页Journal of Systems Science and Mathematical Sciences

基　　金：国家自然科学基金(10271033)资助课题.

摘　　要：本文把文献中关于正态分布下相依回归模型参数Zellner估计的有限样本均方误差 结果和效率结果以及两步协方差改进估计的一般均方误差结果推广到多元t分布情况, 在该分布下两种估计的统计优效性质均不变.Some results of mean square error matrix (MSEM) of the two-stage Aitken estimator introduced by Zellner (1962) in seemingly unrelated regressions (SUR) model with multinormal distribution are extended to the SUR model with a multivariate t-distribution. The exact finite sample result of MSEM of the two-stage covariance-improved estimator introduced by Liu and Wang (1999) is obtained in SUR model with the multivariate t-distribution. It is shown that the finite sample properties of the two-stage Aitken estimator and the covariance-improved estimators are all unchanged under the multivariate t-distribution.

关 键 词：多元T分布 相依回归模型 参数估计 SUR模型 Zellner两步估计 随机矩阵 广义Wishart分布 协方差 Gauss-Markov估计 

分 类 号：O212[理学—概率论与数理统计]