线性模型中回归系数的c-K型估计及其优良性

The c-K Class Estimator of Coefficients in Linear Regression Models and its Optimality

将张建军等人提出的c-k估计推广至c-K估计的情形,证明了在均方误差意义下利用Stein压缩技术可以改进广义岭估计,并给出了下c-K估计优于最小二乘估计的一个充分条件,为病态线性回归模型系数的有偏估计提供了改进的技术途径.

The c-K class estimator proposed by Zhang Jian-jun is extended to the c-k class estimator in this paper and it is proved that the generalized ridge regression estimators can be improved by applying the idea of James-Stein regression technique. Under the mean square error matrix criterion, a sufficient condition on which the c-K class estimator has an advantage over the least square estimator is presented. The method proposed in this paper provides a technical way to the improvement of the biased estimation.