Xiao Huang
 

Department of Mathematics, Washington University
 

Title: Bayesian Mixture Models for Multinomials
 

Abstract: We describe a non-parametric Bayesian model which uses allelic values to genetically classify individuals among populations when the number of overall populations is not specified in advance. It is assumed that each population has a unique set of allele frequencies and a Dirichlet Process is used as the prior distribution for these frequencies, since this process can accommodate a countable number of populations.