Analysis of Bivariate Longitudinal Discrete Data: A Joint Continuous-Time Markov Chains Approach

• Activity Number: 75
• Type: Contributed
• Date/Time: Sunday, August 9, 2015 : 4:00 PM to 5:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Chih-Hsien Wu*

Abstract

In this study, we propose a new methodology for analyzing bivariate longitudinal data with ternary outcomes under a continuous-time Markov chain framework. This method is expected to improve the accuracy of the model estimation to describe the dynamic change of disease by considering two outcome responses simultaneously in the same model. Comparing with previous research, the methodology proposed in this study can deal with Markov chains that have an irreducible infinitesimal matrix and hence will assure the flexibility of disease progression/regression. A shared-parameter introduced in this model will connect the two outcome sequences and be used for describing the correlation between two processes. This methodology also allows the research interests focusing on covariate effects on disease progression, where a log-link function will be used to connect the transition rates and desired covariates. The covariate effects and hence the transition rates will be estimated using Bayesian inference.