Model-free predictor tests in survival regression through sufficient dimension reduction

Abstract

In this article, we test the effects of predictors in survival regression through two well-known sufficient dimension reduction methods. Since the usual sufficient dimension reduction methods do not require pre-specified models, the predictor effect tests can be considered model-free. All of the test statistics have χ 2 distributions. Numerical studies of the proposed predictor effect tests in various simulations and real data application are presented. © 2010 Springer Science+Business Media, LLC.