Title: A recursive estimation of a mixing distribution via a Dirichlet process copula
Authors: Xue Wang - University of Kent at Canterbury (United Kingdom) [presenting]
Abstract: In recent years the mixture of Dirichlet process model in Bayesian nonparametric methods has been analysed most effectively using Markov chain Monte Carlo methods. Recently, Newton and coauthors proposed an alternative fast recursive algorithm for estimation of the mixing distribution which has been shown to perform well in various of applications. However, this algorithm requires evaluation of normalising constants at each iteration. In this talk we will introduce a new recursive scheme via constructing a Dirichlet process copula, which avoids the need for any numerical integration. We illustrate with a number of examples.