Extension of Max-Id Copula to Longitudinal Couple-Based Binary Data

• Activity Number: 175
• Type: Contributed
• Date/Time: Monday, August 2, 2010 : 10:30 AM to 12:20 PM
• Sponsor: ENAR
• Author(s): Seunghee Baek*+
• Companies: University of Pennsylvania
• Keywords: max-id copula ; multivariate ; longitudinal ; binary data

Abstract

This paper presents an extension of max-id copula to longitudinal couple-based data. To allow a complex correlation structure, the conditional distribution of the current observation of each couple is modeled given the previous observation. These conditional distributions are constructed using max-id copula that allows for flexible dependence by using copula parameters. We introduce covariate information into the univariate probability using logistic regression model. We perform simulations, estimating covariate effects on the marginal probabilities and copula parameters, and assessing bias and coverage. The estimates for copula parameters are biased in some scenarios, however, this model performs well in estimating marginal probabilities and correlation structures. Finally, we illustrate this model using data from the Multisite HIV/STD Prevention Trial for African American Couples Study