纵向数据广义部分线性模型的惩罚GMM估计

对纵向数据的部分线性模型,通常的做法是用样条方法或者核方法逼近非参数部分,然后再用广义估计方程的估计方法去估计参数部分.本文使用P-样条拟合非参数函数,对不同的矩条件用不同的广义矩方法对模型的参数和非参数进行估计,并且给出了估计量的大样本性质;并用计算机模拟和实例证明了当模型中存在不同的矩条件时,采用不同的惩罚广义矩方法可以显著地提高估计精度.

Partial Linear Models for Longitudinal Data Based on Penalized General Method of Moments

For the analysis of partial linear modelwith longitudinal data, the general procedure is to fit thenonparametric part with kernel or spline estimation, followed bygeneralized linear model estimating frame. In this paper, we fit thenonparametric part with P-spline, and estimate the parametrical andnonparametric part with different Generalized method of momentsestimation for different moment conditions, implemented by the proofof the asymptotical properties for the estimator, which is also beenproved by simulation and illustrative example, from which we canalso find out that different penalized general method of momentsestimations for different moment conditions perform moreefficiently.