A0551
Title: Describing the dependence structure of clustered right-censored event times through factor copula functions.
Authors: Roel Braekers - Hasselt University (Belgium) [presenting]
Abstract: For clustered right-censored survival data, a copula model is presented to describe the association between different event times within the same cluster. Factor copula functions are used to model the structure within a cluster. This new methodology provides a flexible model for the intra-cluster dependence through the choice of any parametric family of bivariate copulas between the underlying factor and every event time. It is shown that this method extends the existing copula models for this type of data. Furthermore, three estimation procedures in this model are established: a one- and two-stage parametric method and a two-stage semi-parametric method where marginal survival functions are estimated by using a Cox proportional hazards model. For the different parameter estimators, it is proven that they are consistent and asymptotically normally distributed. Through some simulation studies, their finite sample behaviour is assessed. Furthermore, the proposed methods are illustrated on a data set containing the time to first insemination after calving in dairy cattle clustered in herds of different sizes.