Modelos de regresión definidos a trozos con riesgos proporcionales y doble censura

Abstract

The purpose of the present paper is to describe the application of proportional hazards regression models to the analysis of doubly censored survival data, with covariates or risk factors, by means of a parametric modelling of the base-line hazard rate in terms of piecewise distributions. Likelihood equations and associated inferential procedures are presented. Likewise, a consistent estimation of the asymptotic covariance matrix oí the maximum likelihood estimator of the unknown parameter is obtained. In particular, we consider the fitting of the piecewise exponential, Weibull and generalized extreme value distributions. lt can be expected that piecewise models of this kind will usefully describe many proportional hazards survival processes involving changepoints at which the ruling conditions suddenly alter. In addition, severa! methods for evaluating the effects of the risk factors and standard graphical techniques to check model assumptions for uncensored data are modified to allow for double censoring.