Dirichlet process mixture models with shrinkage prior
Abstract
We propose Dirichlet process mixture (DPM) models for prediction and cluster‐wise variable selection, based on two choices of shrinkage baseline prior distributions for the linear regression coefficients, namely, the Horseshoe prior and the Normal‐Gamma prior. We show in a simulation study that each of the two proposed DPM models tends to outperform the standard DPM model based on the non‐shrinkage normal prior, in terms of predictive, variable selection, and clustering accuracy. This is especially true for the Horseshoe model and when the number of covariates exceeds the within‐cluster sample size. A real data set is analysed to illustrate the proposed modelling methodology, where both proposed DPM models again attained better predictive accuracy.Keywords
All Related Versions
Funding Information
- National Science Foundation (SES‐1156372)
This publication has 21 references indexed in Scilit:
- Bayesian nonparametric mixed random utility modelsComputational Statistics & Data Analysis, 2012
- Regression Shrinkage and Selection via The Lasso: A RetrospectiveJournal of the Royal Statistical Society Series B: Statistical Methodology, 2011
- The horseshoe estimator for sparse signalsBiometrika, 2010
- The Bayesian LassoJournal of the American Statistical Association, 2008
- Comparing clusterings—an information based distanceJournal of Multivariate Analysis, 2007
- Slice samplingThe Annals of Statistics, 2003
- 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
- Comparing partitionsJournal of Classification, 1985
- A Bayesian Analysis of Some Nonparametric ProblemsThe Annals of Statistics, 1973