Recursive Polya tree mixture model: a new Bayesian nonparametric model.
Description: We develop a new Bayesian nonparametric model called recursive Polya tree mixture model (RPTMM). This model is very simple and does not require any analytical computation, but can approximate other complicated Bayesian nonparametric approaches. In addition, it is very easy for our model to handle some difficult problems in classical Bayesian nonparametric models. In this talk, we will discuss a baseball player’s data (Brown, 2008), and a rolling thumbtacks data that was originally analyzed through sequential Monte Carlo method (Liu, 1996). We will also compare our RPTMM to a meta-analysis data with a complicated conditional Dirichlet process (Burr and Doss, 2005), and develop a Bayesian semiparametric AFT model (Hanson and Johnson, 2004). This talk is based on the joint work with professors David Stephens and James Hanley.
Li Shujie is a PhD (Biostatistics) student at McGill University, working under the supervision of Drs. James Hanley and David Stephens. His research interest is to develop flexible Bayesian nonparametric and semi-parametric models for medical applications, including random-effects meta-analysis, Bayesian semi-parametric accelerated failure time model for survival data, and recurrent data analysis.