摘要
主要研究因变量存在缺失且协变量部分包含测量误差情形下,如何对变系数部分线性模型同时进行参数估计和变量选择.我们利用插补方法来处理缺失数据,并结合修正的profile最小二乘估计和SCAD惩罚对参数进行估计和变量选择.并且证明所得的估计具有渐近正态性和Oracle性质.通过数值模拟进一步研究所得估计的有限样本性质.
This paper focuses on simultaneous parameter estimation and variable selection to varying-coefficient partially linear models when covariates of parametric component are measured with additive errors and responses are missing. The imputation method is applied to handle missing responses, and a variable selection procedure is proposed by using corrected profile least-squares method and SCAD penalization, and the resulting estimators perform asymptotic normality as well as the oracle property. Simulation studies are conducted to illustrate the finite sample properties of the proposed procedures.
出处
《数学的实践与认识》
CSCD
北大核心
2014年第16期122-128,共7页
Mathematics in Practice and Theory
基金
中央高校基本科研业务费专项资金资助(ZZ1229)
关键词
变系数部分线性模型
缺失数据
测量误差
SCAD
变量选择
varying coefficient partially linear models
missing data
measurement error SCAD
variable selection
