Smoothing parameter selection in semiparametric regression models with censored data

Abstract

In this paper, we introduce penalized spline estimators for the unknown function and a parameter vector in a semiparametric regression model with right censored data. In order to obtain this estimator accurately and efficiently, we used penalized spline method based on three important selection criteria such as corrected Akaike's information criterion (AIC), generalized cross-validation (GCV) and Mallows' Cp criterion (MCp). The purpose of the study is to illustrate the performance of penalized regression spline method for estimating the right-censored data and also comparing the mentioned three selection methods in selection of smoothing parameter. The ideas that expressed in this study are demonstrated in a real cancer patients' data and a Monte Carlo simulation using different censoring levels and sample sizes. Thus, the appropriate selection criteria are provided for a suitable smoothing parameter selection. Cp gave satisfying results for this study. (C) 2017 The Authors. Published by IASE. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).