B1656
Title: Bayesian clustering factor models
Authors: Allison Tegge - Virginia Tech (United States) [presenting]
Marco Ferreira - Virginia Tech (United States)
Hwasoo Shin - Virginia Tech (United States)
Abstract: A novel Bayesian factor model is proposed to cluster multivariate data. The motivating example is in the area of recovery from substance use disorder. The model assumes that the common factors follow a mixture of Gaussian distributions. The particular interest is in correctly estimating the number of components in the mixture, which is known to be a non trivial problem. For this challenging model selection problem, a unit-information prior is developed. In addition, a Markov chain Monte Carlo algorithm is developed for posterior exploration. Results are presented from a simulation study comparing different methods for choosing the number of components in the mixture. The usefulness and flexibility are illustrated of the proposed approach with an application to recovery from substance use disorder. Together, the number of components in the mixture of Gaussians and the characteristics of the subjects in each group have implications for medical treatment.
