Nonparametric Bayes applications to biostatistics

David Dunson

Department of Statistical Science, Duke University

March, 2008

This article provides a review of recent developments in nonparametric Bayes modeling in biostatistical applications, with a focus on flexible hierarchical models with nonparametric priors for random effects distributions. The emphasis is on practical advantages of nonparametric Bayes methods and issues that arise in prior specification and computation. Some topics covered include Dirichlet process mixture models, methods for functional data analysis and new approaches that allow distributions to change flexible with time, predictors and spatial location. Several applications to epidemiologic studies are included.

Keywords: Dirichlet process; Functional data; Hierarchical modeling;Mixture model; Random effects; Semiparametrics

The manuscript is available in PDF formats.