Skip to content
Prev 721 / 20628 Next

lme4::mcmcsamp + coda::HPDinterval

Hi, Prof. Bates,

Douglas Bates said the following on 4/1/2008 11:40 AM:
 > This is a known problem with MCMC sampling of
 > variance components.

So, does this mean the confidence interval is in question, and not the 
point estimate? Would Wald intervals not be appropriate either? Or 
profile likelihood intervals? (I know neither are available in lme4 but 
I think could hack it if necessary.)

I've updated the lme4 package have started testing it. Thanks also for 
the Gelman reference. I've requested a copy and will take a look.

Thanks,

--sundar

Douglas Bates said the following on 4/1/2008 11:40 AM: