I have just uploaded a new version of the lme4 package to R-forge.R-project.org. I hope that binary packages for lme4_0.999875-6 will be available tomorrow. This version has an mcmcsamp function for linear mixed models fit by lmer. I have not yet incorporated the Metropolis-Hastings step for sampling from the fixed effects and random effects in generalized linear mixed models or nonlinear mixed models. I have methods for the HPDinterval function for these objects. You can apply it to the entire sample or to the fixef slot only or to the sigma slot only. There is another slot called ST from which the variances and covariances of the random effects can be derived. I think I found a glich in earlier versions of mcmcsamp that made them overly conservative. When I sampled from the conditional distribution of the sigma parameter I think I used the wrong number of degrees of freedom for the chi-square distribution. I have modified that and think the results look more like they should. However, I would appreciate feedback on whether these results look sensible to others.
New alpha version of lme4 code on R-forge
1 message · Douglas Bates