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

作　　者：许晓丽[1] 赵明涛[2] Xu Xiaoli;Zhao Mingtao(School of Management Science and Engineering,Anhui University of Finance and Economics,Bengbu Anhui 233030,China;Institute of Statistics and Applied Mathematics,Anhui University of Finance and Economics,Bengbu Anhui 233030,China)

机构地区：[1]安徽财经大学管理科学与工程学院,安徽蚌埠233030 [2]安徽财经大学统计与应用数学学院,安徽蚌埠233030

出　　处：《统计与决策》2022年第19期10-15,共6页Statistics & Decision

基　　金：国家社会科学基金青年项目(15CTJ008);安徽省高校人文社会科学研究项目重点项目(SK2020A0051)。

摘　　要：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.

关 键 词：纵向数据 测量误差数据 可加偏线性测量误差模型 二次推断函数 惩罚二次推断函数 

分 类 号：O212.7[理学—概率论与数理统计]