We also compute credible intervals for the random effects variances; whereas, there is no good approach for estimating uncertainty for PQL [penalized quasi-likelihood] variance estimates, or other frequentist variance estimates. In addition we are able to simultaneously compute marginal posterior inclusion probabilities for both the ﬁxed effects and random effects and correctly locate the true model as the one with highest posterior probability.
Next on my list of things to code
Kinney, S. K. and D. B. Dunson (2007), “Fixed and random effects selection in linear and logistic models,” Biometrics (63), 690–698. (PDF; R code)