回归系数地混合估计和最小二乘估计地相对效率.docx

山东师范天学硕lt学位论文 ．!∑(qyil+盯!瓯)“ e5≥P 4上L_—————～ 25一，+；o-1-P ’ 关键词：线性回归模型， 混合估计， 最小二乘估计， 相对效率 协方差矩阵，上界，下界． 分类号：0212．1 山东师范人学坝l‘学位论文 The RelatiMe ElfiCiencies of Mi xed Estimator With respect to Least Square Estimator in Linear Regression Mode Jiafu Yu Shandong Normal University，Jinan，Shandong， 250014 People’S Republic of China ABSTRACT The relative efficiencies of parameter estimators in linear model are the problems which we talk about often． Usually the parameter est．imators have two kinds，one iS the best】inear unbiased estimator of the unknown parameter(BLUE)，and the other is the linear square est imator of the unknown parameter(LSE) First we consider the following models： Y=Xfl+e，E(0=o，Cov(e)=盯2I， (1) Y=Xfl+e，E(e)=o，Cov(e)=盯2V． (2) ’莲nmodel(1)，we know that the best linear unbiased estimator of function C’口iS equal to its linear square estimator，hut in many questions，we ca／3’t think that the hypothesiS of the error vectors i S right．Because their deviation may be not equal to each Other，and maybe CO{、Fefati to each other，at this time，its covariance iS Cov(e)=盯2V，and the model becomes model(2)，in thiS model，when the matrix has complete rank，∥ is estimating，and its least square estimator is口=(xx)～XT，its best linear unbiased estimator iS声+=(Ⅳ’V“x)“Z矿一Y，but in many theory and utility prob]ems，V often includes unknown parameters．If we wrfte the 6 坐查壁婆盔兰竺!．兰些堕苎 unknown parameters 0，we should use the estimator of口to take place 0 j n口+，and this lead t,o two steps estimator．But it iS a pity that ate not fami 1 i￡lr wi th the study of two steps estimators’ statistical properties，and sometimes we can’t obtain any estimator of咿，SO we turn to use least square estimator彦，when we use least square estimator to take place best 1inear unbiased estimator，we may sustain some loss．and sometimes the loss can become very large．I L i S very important to study this kind of loss．So people proljosg‘+the!de}ibition of relat