摘要
本文采用广义岭估计β(K)来估计增长曲线模型中回归系数β=vec(B),通过K值的选取,可使其均方误差(MSE)小于LS估计β的MSE。同时对LS估计的任一线性变换,给出了其均方误差的一个无偏估计,并应用极小化β(K)的MSE的无偏估计的方法,得到了确定岭参数的公式。
In this paper, the generalized Ridge estimator β(K) of regression coffi-cients β = Vec(B) is considered in growth curve model. The mean square error(MSE) of above estimate is less than MSE of LS estimate β by choosing the matrix K(Ridge paremeter). In the meantime, the unbiased estimate of MSE of any linear tranformation of LS estimate β is given, and the formula of determing Ridge parameter is derived by the method of minimized unbiased estimate of MSE.
关键词
增长曲线模型
岭估计
均方误差
广义
growth curve model, Ridge estimate, mean square error.