Message-ID: <40e66e0b0807110719w7da068cyb7d74dde8b6dc05d@mail.gmail.com>
Date: 2008-07-11T14:19:44Z
From: Douglas Bates
Subject: Covariance structure and MCMC
In-Reply-To: <DE0D5C6148C2F24B9CCE3F10DE2AF00AC42AF2@ut40se02.atk.com>
On Fri, Jul 11, 2008 at 9:06 AM, Remund, Todd <Todd.Remund at atk.com> wrote:
> I have been using lmer() to fit a mixed model to some experimental design
> data. I have an unstructured covariance structure. I would like to use
> your mcmcsamp() function to to inference using a Bayesian methods, but the
> mcmcsamp function will not take the mer class object when it has the
> covariance structure. Is there a prototype version that I could use to
> create the MCMC samples? I really appreciate your time, and the functions
> you have created. They are very fast. (My data set is rather large.)
Thank you for the kind words. I'm sorry to say that I don't have a
prototype version of mcmcsamp for models with correlated random
effects. I believe it would take only a day or two to implement it
but right now I don't have a day or two to work on it. There are
simply too many other commitments at this time, including checking on
the peculiar behavior of the current mcmcsamp for models with
uncorrelated random effects. There is really no purpose in working
out the sampling of the parameters that induce correlation if the
variance components are not being sampled correctly.