A Split-Merge Markov chain Monte Carlo Procedure for the Dirichlet Process Mixture Model
- 1 March 2004
- journal article
- Published by Taylor & Francis Ltd in Journal of Computational and Graphical Statistics
- Vol. 13 (1), 158-182
- https://doi.org/10.1198/1061860043001
Abstract
. We propose a split-merge Markov chain algorithm to address the problem of inefficientsampling for conjugate Dirichlet process mixture models. Traditional Markov chain MonteCarlo methods for Bayesian mixture models, such as Gibbs sampling, can become trapped in isolatedmodes corresponding to an inappropriate clustering of data points. This article describes aMetropolis-Hastings procedure that can escape such local modes by splitting or merging mixturecomponents. Our Metropolis-Hastings...Keywords
This publication has 13 references indexed in Scilit:
- Modelling Heterogeneity With and Without the Dirichlet ProcessScandinavian Journal of Statistics, 2001
- Computational and Inferential Difficulties with Mixture Posterior DistributionsJournal of the American Statistical Association, 2000
- Simulating normalizing constants: from importance sampling to bridge sampling to path samplingStatistical Science, 1998
- A semiparametric Bayesian model for randomised block designsBiometrika, 1996
- Bayesian Density Estimation and Inference Using MixturesJournal of the American Statistical Association, 1995
- Estimating Normal Means with a Dirichlet Process PriorJournal of the American Statistical Association, 1994
- Estimating normal means with a conjugate style dirichlet process priorCommunications in Statistics - Simulation and Computation, 1994
- Mixtures of Dirichlet Processes with Applications to Bayesian Nonparametric ProblemsThe Annals of Statistics, 1974
- Ferguson Distributions Via Polya Urn SchemesThe Annals of Statistics, 1973
- Monte Carlo sampling methods using Markov chains and their applicationsBiometrika, 1970