MCP-penalized Regression in High Dimensional Partially Linear Models for Right Censored Data

Abstract

Basing on Stute’s weighted least squares method for prognosis studies with right censored response variables, this paper first put forward a semi-parametric regression model include two covariate effects, one is for the low dimensional covariates taken a nonparametric form, the other is for the high dimensional covariates taken a parametric form. Then, the selection of parametric covariate effects was achieved by use of a minimax concave penalty (MCP) approach, it is with great consistency in the semi-parametric regression model. In succession, the nonparametric component was estimated by a sieve approach. Finally, a numerical simulation was made, the result shows that the proposed approach has satisfactory performance.