最优线性无偏估计和Bayes估计的优良性研究.pdfVIP

Research on th e problem s of B L U estim ator and Bayes estim ator in linear m odel A bstract In this paper, we study the superiority of best linear unbiased (BLU) estimator and Bayes estimator in the linear model. The paper is divided into four parts. In the first part, we make a brief review about the development in the theory of BLU estimator and Bayes estimator in the linear model, and also introduce some matrix theory related to this paper. In the second part, we study the robustness of BLU estimator in the linear model with respect to a linear equality constraint. The suffi­ cient and necessary conditions for the BLU estimator of the estimable function C^ to be robust are obtained. In the third part, we investigate a general class of linear model with an equality constraint. Without specifying the distribution of the model error, we obtain the necessary and sufficient conditions for the existence of BLU estimators under quadratic loss and matrix loss. In the fourth part, the superiority of the Bayes linear unbiased estimators of unknown parameters in a class of linear models is studied. We obtain the conditions of the Bayes linear unbiased estimators to be superior to the generalized least squares estimator under balanced loss criterion and MSEM criterion. Keywords: equality constraint; BLU estimator; Bayes linear un­ biased estimator; Robustness; Balanced loss ; MSEM criterion; Superi­ ority. 万方数据 目 录 § 1 引言及预备知识 1 §1.1 引言 1 §1.2