A Bayesian Approach for the Analysis of Latent Stage-Sequential Process: Timing of Drinking Onset and Sequential Patterns of Alcohol Use

• Activity Number: 603
• Type: Topic Contributed
• Date/Time: Thursday, August 4, 2011 : 8:30 AM to 10:20 AM
• Sponsor: Section on Health Policy Statistics
• Author(s): Hwan Chung*+
• Companies: Ewha Womans University
• Keywords: early drinking onset ; label switching ; latent class profile analysis ; MCMC ; stage-sequential process ; under-age drinking

Abstract

Research on the sequential development of alcohol use can be challenging in part because the stage of drinking behavior at a certain time point may not be directly observable. In this context, a latent class analysis (LCA) can provide a set of principles for systematic identification of homogeneous subgroups. We propose an LCA approach, referred to as a latent class-profile analysis (LCPA), for investigating the dependency of stage-sequential patterns of adolescent drinking behavior on timing of drinking onset. Analyses indicate that early drinking onset is related to increased alcohol use. Among adolescents aged 12-14, those who already have tried alcohol are more likely to become heavy drinkers compared to their late-onset counterparts. Parameters for LCPA are estimated by Bayesian method via Markov chain Monte Carlo (MCMC). This paper proposes a dynamic-dependent prior for LCPA to deal with label switching. We argue that when conventional Bayesian estimates behave erratically, problems often may be alleviated with a small amount of prior input for the LCPA model.