Bayesian DP Time-Varying Circumplex Model

• Activity Number: 621
• Type: Contributed
• Date/Time: Thursday, August 2, 2012 : 8:30 AM to 10:20 AM
• Sponsor: Statistics in Mental Health Research
• Author(s): Zhaowei Hua*+ and David Dunson and Sy-Miin Chow and Hongtu Zhu
• Companies: Millennium Pharmaceuticals, Inc. and Duke University and The University of North Carolina at Chapel Hill and The University of North Carolina at Chapel Hill
• Keywords: circumplex correlation ; emotion dynamics ; Dirichlet process ; ordinal scales ; Bayesian

Abstract

In the analysis of longitudinal circumplex data, it is of interest to explore the dynamic transition pattern of affective experience. The circumplex model of affect postulates that the underlying structure of affective experience can be characterized as an ordering of affective states on the circumference of a circle. This article proposes a semiparametric Bayesian DP time-varying circumplex model. A Dirichlet proces is used to capture the latent clustering structure and a discrete time-varying circumplex model addresses the time dependent transition structure and the self-transition feature. We develop an efficient MCMC algorithm using the exact block Gibbs sampler. The performance of our method is assessed by a simulation study. We apply our method to analyze a data set of emotion ordinal scales from a clinical study of the dynamics of emotion regulation.