摘要
本文对多元线性模型回归系数提出了主成分估计，并证明了主成分估计优于最小二乘估计。进一步，对最小二乘估计的任一线性变换，给出了均方误差的一个无偏估计，并应用极小化均方误差的无偏估计的方法，给出了确定偏参数的公式。
Abstract This paper considers the principal components estimate of regression coefficient in multivariate linear models and proves that the principal cimponents estimate is superior to least square(LS) estimate.Furthermore,the unbiased estimate of mean square error(MSE) of any linear transformation for LS estimate is given. Using the method of minimizod unbiased estimate of MSE,the formula of determining biased parameter with the principal components estimate is shown.
出处
《工程数学学报》
CSCD
1994年第2期100-104,共5页
Chinese Journal of Engineering Mathematics
