误差方差阵的二次型的经验Bayes估计的收敛速度

THE CONVERGENCE RATE OF EMPIRICAL BAYES ESTIMATION FOR THE QUADRATIC FORM OF ERROR COVARIANCE MATRIX

经验Bayes(EB)估计是一种有效的参数估计,其收敛速度问题已受到人们的广泛注意,文献[1～3]都曾在这方面作过一些工作。本文考虑如下线性模型:

In this paper, empirical Bayes (EB) estimation for the quadratic form l′∑~1l of inverse of error covariance matrix ∑ in normal linear model is proposed. We prove that the convergence rate of EB estimation about l'∑^(-1)l can be arbitarily close to 1. Finally, an example is given.