A General Semiparametric Bayesian Model and Software Package for Multistate Data

• Activity Number: 601
• Type: Contributed
• Date/Time: Wednesday, August 12, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Adam King* and Robert E. Weiss
• Companies: California Polytechnic State University and UCLA
• Keywords: Software ; Survival ; Multistate ; MCMC ; Competing Risks ; Semiparametric

Abstract

Multistate data consists of records of time points at which subjects make transitions among a discrete collection of states. We propose simultaneously modeling the hazards of all possible transitions between states by relating the hazards of each transition type to linear predictors containing additive functions of arbitrary numbers of time-varying covariates. With our software, Bayesian inference for this model is automated via a Block full-conditional-approximating Metropolis-Hastings routine, which provides good MCMC mixing without the need for adaptation or user-tuning. We demonstrate our model and software using data on recurrent episodes of drug use and related traits in a population of illicit drug users.