Partially Linear Additive Hazards Regression With Varying Coefficients

Abstract

To explore the nonlinear interactions between some covariates and an exposure variable, we propose the partially linear additive hazards model for survival data. In a semiparametric setting, we construct a local pseudoscore function to estimate the varying and constant coefficients and establish the asymptotic normality of the proposed estimators. Moreover, we develop the weak convergence property for the local estimator of the baseline cumulative hazard function. We conduct simulation studies to empirically examine the finite-sample performance of the proposed methods and use real data from a breast cancer study for illustration. © 2008 American Statistical Association.