EcoSta 2017: Start Registration
View Submission - EcoSta2017
A0481
Title: A Bayesian joint frailty-copula approach for modeling recurrent events and a terminal event Authors:  Ming Wang - Pennsylvania State University (United States) [presenting]
Abstract: In practice, recurrent events are always encountered, but sometimes will be censored by a terminal event. The non-informative censoring assumption is violated under this situation, and as a result, we cannot model the recurrent event process alone. The joint frailty model is widely used to jointly model these two processes. However, there exist limitations due to the assumption of conditional independence given a subject-level frailty and indirect estimate of their association. Besides this, Copula is a popular approach to model bivariate time to event processes. In order to relax the conditional independence assumption and estimate the association directly, we propose a joint frailty-copula approach under a Bayesian framework to model the terminal time-to-event process and the recurrent time-to-event process. We show that the joint frailty-copula model is a more generalized model, extended from a nested frailty model. Metropolis-Hastings within Gibbs Sampler algorithm is used for estimating parameters. Extensive simulation studies are performed to evaluate the performance of our method in terms of bias, mean squared error and robustness. Finally, we apply our method to analyze the Marketscan data, studying the association between recurrent stroke process and the death process.