Chapter

Bayesian Variable Selection for Random Intercept Modeling of Gaussian and Non‐Gaussian Data

Sylvia Frühwirth‐Schnatter and Helga Wagner

in Bayesian Statistics 9

Published in print October 2011 | ISBN: 9780199694587
Published online January 2012 | e-ISBN: 9780191731921 | DOI: http://dx.doi.org/10.1093/acprof:oso/9780199694587.003.0006
Bayesian Variable Selection for Random Intercept Modeling of Gaussian and Non‐Gaussian Data

More Like This

Show all results sharing this subject:

  • Probability and Statistics

GO

Show Summary Details

Preview

The paper demonstrates that Bayesian variable selection for random intercept models is closely related to the appropriate choice of the distribution of heterogeneity. If, for instance, a Laplace rather than a normal prior is considered, we obtain a Bayesian Lasso random effects model which allows both smoothing and, additionally, individual shrinkage of the random effects toward 0. In addition, we study spike‐and‐slab random effects models with both an absolutely continuous and a Dirac spike and provide details of MCMC estimation for all models. Simulation studies comparing the various priors show that the spike‐and‐slab random effects model outperforms unimodal, non‐Gaussian priors as far as correct classification of non‐zero random effects is concerned and that there is surprisingly little difference between an absolutely continuous and a Dirac spike. The choice of appropriate component densities, however, is crucial and we were not able to identify a uniformly best distribution family. The paper concludes with an application to ANOVA for binomial data using a logit model with a random intercept.

Keywords: Bayesian Lasso; MCMC; spike‐and‐slab priors; shrinkage

Chapter.  19434 words.  Illustrated.

Subjects: Probability and Statistics

Full text: subscription required

How to subscribe Recommend to my Librarian

Buy this work at Oxford University Press »

Users without a subscription are not able to see the full content. Please, subscribe or login to access all content.