多元正态分布均值参数估计的Γ-非容许性

On the F-inadmissibility of the Estimator of a Multivariate Normal Mean

本文考虑了正态分布均值参数估计的Γ-非容许性。首先给出了微分不等式△(x)=Ψ(x)sum from i=1 to p(v_i(x_i)(?)_i^(i(1))(x))+sum from i=1 to p(w_i(x)(?)_i^2(x))<0较广的一类解,然后证明了James-Stein型估计量是Γ-非容许的,并给出了改进估计量。

The present paper describes the Γ-inadmissibility problem of estimation of a multivariate normal mean. A new solution to the differential inequality is given and it is proveed that the James and Stein type estimator δ^(Js) is Γ-inadmissible. Thus an explicit improved estimator is given.