Slice Sampling Mixture Models

We propose a more efficient version of the slice sampler for Dirichlet process mixture models described by Walker (2007). This sampler allows the fitting of infinite mixture models with a wide–range of prior specification. To illustrate this flexiblity we develop a new nonparametric prior for mixture models by normalizing an infinite sequence of independent positive random variables and

show how the slice sampler can be applied to make inference in this model. Two submodels are studied in detail. The first one assumes that the positive random

variables are Gamma distributed and the second assumes that they are inverse–Gaussian distributed. Both priors have two hyperparameters and we consider

their effect on the prior distribution of the number of occupied clusters in a sample. Extensive computational comparisons with alternative ”conditional” simulation techniques for mixture models using the standard Dirichlet process prior and our new prior are made. The properties of the new prior are illustrated on a density estimation problem.