Analysis of Bivariate Survival Data Based on Copulas with LogGEV Marginals

• Activity Number: 193
• Type: Contributed
• Date/Time: Monday, August 10, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Dooti Roy* and Vivekananda Roy and Dipak K. Dey
• Companies: University of Connecticut and Iowa State University and University of Connecticut
• Keywords: Bivariate Data ; Survival Analysis ; Copula ; Generalized Extreme Value Distribution ; Diabetes Retinopathy ; Empirical Bayes

Abstract

During the last two decades, there have been a growing interest in modeling multivariate survival data. It has been medically observed that incidence of one disease often increases the risk of another in a patient as in HIV patients. We model bivariate survival data using copula, and in this paper, our goal is to develop the Bayesian framework for inference and estimation of the marginals and the dependence parameter following the approach outlined in Roy (Computational Statistics and Data Analysis, 2014) using Clayton copula structure and the flexible GEV distribution to model the marginal distributions. We propose an empirical Bayes approach based on importance sampling method. It is computationally efficient as one has to use only one MCMC chain to estimate the dependency parameter. We additionally present simulations and a real data application to demonstrate the applicability of our method.