相依线性回归系统的压缩估计

Shrunken estimators of seemingly unrelated regression model

对于相依线性回归系统(SUR,m=2),其设计阵为病态时,文中给出了βi(i=1,2)的压缩估计及相应的两步估计,证明了在θi(i=1,2,…,p1)满足一定条件时,此估计相对于协方差改进估计及两步协方差改进估计的优良性,同时给出了θi(i=1,2,…,p1)的确定方法.

For seemingly unrelated regression equations with design matrices being ill-conditioned,we proposed the shrunken principal estimators and the related two-stage estimators of βi（i=1,2）.When θi（i=1,2,…,p1） satisfied a certain condition,we proved the superiority to CI estimator and two-stage CI estimator.Finally,we gave a method to find the solution of θi（i=1,2,…,p1）.