纵向数据下可加偏线性测量误差模型的变量选择

Variable Selection of Additive Partial Linear Measurement Error Model Based on Longitudinal Data

文章基于纵向数据研究可加偏线性测量误差模型的模型选择,提出了一种用于模型估计和选择的惩罚二次推断函数方法。利用该方法得到的非零参数的估计是相合的、渐近正态的,可加函数的估计具有最优收敛速度。数值模拟结果显示,在有限样本情况下,该方法要优于基于LASSO惩罚函数的惩罚二次推断函数方法。

This paper studies the model selection of additive partial linear measurement error model based on longitudinal data,and proposes a penalized quadratic inference function method for model estimation and selection.The estimators of nonzero parameters obtained by use of this method are consistent and asymptotically normal,and the estimators of additive functions have the optimal convergence rate.Numerical simulation results show that the proposed method outperforms the penalized quadratic inference function method based on LASSO penalty function in the case of finite samples.