Bayesian Analysis of Longitudinal Binary Data Using Multivariate Bridge and Other Random Effects Models

• Activity Number: 629
• Type: Invited
• Date/Time: Thursday, August 2, 2012 : 10:30 AM to 12:20 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Bani K Mallick*+ and Souparno Ghosh and Debajyoti Sinha and Stuart Lipsitz
• Companies: Texas A&M University and and Florida State University and Harvard Medical School
• Keywords: orrelated longitudinal binary data ; multivariate normal distributio ; partial linear model ; probability integral transformation ; MCMC

Abstract

We present Bayesian methods of analysis for longitudinal binary outcomes using the random effects models where the marginal link function, when integrated over the distribution of the random effects, is of the same form as that of the conditional link. We propose novel models for longitudinal data, with separate, but correlated, random effects with multivariate bridge and positive stable distributions. The proposed copula model allows the marginal correlation among the binary outcomes within same subject to decline with increasing time separation while retaining the same form of conditional and marginal link functions. Our models and associated methodologies have been illustrated with the analysis of a longitudinal binary data from an AIDS study.