纵向缺失数据下高维部分线性回归模型的变量选择  

Variable Selection for High-dimensional Partially Linear Regression Models with Longitudinal Missing Data

作　　者：田瑞琴 徐登可 TIAN Ruiqin;XU Dengke(School of Science,Hangzhou Normal University,Hangzhou 311121,China;School of Science,Zhejiang Agriculture and Forestry University,Hangzhou 311300,China)

机构地区：[1]杭州师范大学理学院,浙江杭州311121 [2]浙江农林大学理学院,浙江杭州311300

出　　处：《杭州师范大学学报（自然科学版）》2020年第3期273-281,共9页Journal of Hangzhou Normal University(Natural Science Edition)

基　　金：国家自然科学基金项目(11801514);浙江省自然科学基金项目(LY17A010026).

摘　　要：针对纵向单调缺失数据下高维部分线性回归模型的变量选择,提出了逆概率加权光滑阈估计方程变量选择方法,其中非参数分量使用样条估计.在适当正则条件下,证明了该变量选择方法具有Oracle性质,并通过模拟研究验证了所提出方法的有限样本性质.For the variable selection in high-dimensional partially linear models with longitudinal missing data,a variable selection method based on an inverse probability weighted smooth-threshold estimation equation is proposed,in which the nonparametric component is estimated by splines.Under some regularity conditions,the variable selection method is proved to have Oracle property.Finally,simulation studies are conducted to examine the finite sample performance of the proposed method.

关 键 词：纵向数据 单调缺失 变量选择 广义估计方程 B样条 

分 类 号：O212.7[理学—概率论与数理统计]