摘要
一般得到了两方程相依回归模型的任一线性估计在均方误差准则下优于最小二乘估计的充要条件,据此提出一种新的广义非限定两步估计类(非线性),推导得到了这种两步估计的精确均方误差结果,研究了它优于最小二乘估计,甚至优于Zellner估计的统计性质.
For a system of two seemingly unrelated regression equations, a necessary and sufficient condition is obtained for a general linear estimator of the regression coefficients to be better than the LS estimator in terms of the mean square error matrix (MSEM) criterion. According to this result a class of two-stage estimators based on a generalized unrestricted estimate of the dispersion matrix is suggested. The exact MSEM of the two-stage estimator is derived,its superiorities over the LS estimator and even over the Aitken estimator introduced by Zellner (1963) are examined. An F-distribution statistic for testing MSEM-dominance is provided.
出处
《五邑大学学报（自然科学版）》
CAS
2002年第3期10-14,共5页
Journal of Wuyi University(Natural Science Edition)
