正规矩阵R及平滑因子α的选取

Choice of the regular matrix R and smoothing parameter α

文献1对半参数模型yi=XTiβ+s(ti)+ei(i=1,2,…,n)运用补偿最小二乘法则VTPV+αS^TRS^=min,有效地对β和s(ti)进行了估计,但该文没有解决平滑因子α和正规矩阵R的选取问题。本文对此进行了初步的研究,得到和总结了选取α和R的一些方法,以供测量数据处理者参考。

In literature, for the semiparametric model y=Xa+s(t)+e(i=1,2,...,n), the estimations of parameter a and nonparameter s(t) are effectively obtained using the method of penalized least squares VPV+á R=min,, but choice of the regular matrix R and smothing parameter á are not solved. In the paper, the question is elementarily investigated. Moreover, some methods of choice are obtained and summarized, the researchers who process measuring data can make reference.