[R-sig-ME] model selection with mcmc

[catferr at libero.it]

Dear all
  I am trying to run some models on behavioral data by using lme4 and 
mcmcsamp. I am new with this approach and  I have a question I would like to 
ask you: once I select the best random structure, what is the best way (or more 
advisable at this stage) to do a model selection from a full model (e.g. 
including all potential important variables as fixed effects) to the best 
model? I was wondering if it would be better to use the AIC criterium or to run 
the mcmcsamp on the full model. in case I use the mcmcsamp on the full model 
which is the correct way to proceed? I will check for further reference on this 
topics but if you have some suggestion it would be very appreciated.

thanks a lot
Caterina Ferrari
UQAM