A MODELLING APPROACH TO INCREASE THE EXPLAINED RISK IN THE PROPORTIONAL HAZARDS REGRESSION

Abstract

In this study, a modelling strategy is developed to obtain more information from censored obser-vations. By the proposed approach, uncensored observations are clustered using a fuzzy c-means algorithmand the degrees to which censored observations are members of these clusters are determined. Censoredobservations are weighted based on their membership values and the distances between the censoring timeand the time components of the cluster centres. Further, simulation studies are performed to characterizethe performance of the proposed approach based on the explained risk measure.