Variable screening for varying coefficient models with ultrahigh-dimensional survival data

Abstract

In this article, we develop a variable screening method for varying coefficient hazards models of single-index form. The proposed method can be viewed as a natural survival extension of conditional correlation screening. An appealing feature of the proposed method is that it is applicable to many popularly used survival models, including the varying coefficient additive hazards model and the varying coefficient Cox model. The proposed method enjoys the sure screening property, and the number of the selected covariates can be bounded by a moderate order. Simulation studies demonstrate that our method performs well, and an empirical example is also presented.