Abstract:

Nonparametric Bayesian methods need to construct a prior measure on the space of random distributions with large support. The Dirichlet process and mixtures of Dirichlet processes have been popular prior measures used in machine learning, biostatistics, bioinformatics, etc. In this talk, I will explain both Dirichlet processes and mixtures of Dirichlet processes and their usages in many biostatistics problems. These problems include functional analysis of gene expression data for time course experiments, quantal bioassay, and random effects models.