Two-level stochastic search variable selection in GLMs with missing predictors.

Published

Journal Article

Stochastic search variable selection (SSVS) algorithms provide an appealing and widely used approach for searching for good subsets of predictors while simultaneously estimating posterior model probabilities and model-averaged predictive distributions. This article proposes a two-level generalization of SSVS to account for missing predictors while accommodating uncertainty in the relationships between these predictors. Bayesian approaches for allowing predictors that are missing at random require a model on the joint distribution of the predictors. We show that predictive performance can be improved by allowing uncertainty in the specification of predictor relationships in this model. The methods are illustrated through simulation studies and analysis of an epidemiologic data set.

Full Text

Duke Authors

Cited Authors

  • Mitra, R; Dunson, D

Published Date

  • January 2010

Published In

Volume / Issue

  • 6 / 1

Start / End Page

  • Article - 33

PubMed ID

  • 21969986

Pubmed Central ID

  • 21969986

Electronic International Standard Serial Number (EISSN)

  • 1557-4679

International Standard Serial Number (ISSN)

  • 2194-573X

Digital Object Identifier (DOI)

  • 10.2202/1557-4679.1173

Language

  • eng