联立方程模型参数的主成分压缩估计  

On Principal Component Shrinkage Estimates of Parameters in Multicollinear Simultaneous Equation Models

作　　者：胡俊航[1] 童恒庆[2] HU Jun-hang;TONG Heng-qing(Centre for Quality Science Research,Henan Quality Polytechnic,Pingdingshan Henan 467001,China;School of Sciences,Wuhan University of Technology,Wuhan 430070,China)

机构地区：[1]河南质量工程职业学院质量科学研究中心,河南平顶山467001 [2]武汉理工大学理学院,武汉430070

出　　处：《西南师范大学学报（自然科学版）》2018年第3期15-20,共6页Journal of Southwest China Normal University(Natural Science Edition)

基　　金：国家自然科学基金项目(81271513;91324201)

摘　　要：在外生变量设计矩阵X复共线时,证明了恰好识别的联立方程模型参数的间接最小二乘估计的均方误差变大的事实,为改善间接最小二乘估计的偏误,提出了参数的主成分压缩估计,并讨论了其优良的统计性质,数值模拟也验证了主成分压缩估计优于间接最小二乘估计.If matrix of the exogenous variables is multicollinear in identified simultaneous equations,the fact that mean square error of the indirect least squares estimates of parameters is bigger has been proved.In order to improve the deviation of parameters estimated by indirect least squares,the principal component shrinkage estimates of the parameters have been proposed,and its excellent statistical properties been discussed.Numerical simulation validates that the principal component shrinkage estimate is more excellent than the indirect least squares estimate.

关 键 词：联立方程模型 参数估计 间接最小二乘估计 主成分压缩估计 

分 类 号：O212[理学—概率论与数理统计]