[R-sig-ME] lme4 mcmcsamp workaround?

Michael Erickson erickson at ucr.edu
Tue Sep 8 17:36:59 CEST 2009

I have tried searching the archives to find an answer to this
question, but I'm not having much luck.  What I think I have
discovered is that since the release of lme4_0.999375, mcmcsamp() has
not worked for correlated random effects.

I looked at the r-forge site and it seems like this is bug [#68]
"mcmcsamp is broken."  The last update on it was 2008-06-23, and it
did not sound promising.

What are people doing now without mcmcsamp()?  Are they not getting
HPDintervals for correlated random effects?  Are they only using
models with uncorrelated random effects? Have they reverted to a
version that works in some cases, but not when values of a variance
component are near 0?

What I am wondering, is whether or not my life is better not being
able to do this or whether I can use some sort a workaround.  If there
is a workaround, can someone tell me what it is?  I am currently using
lme4_0.999375-31 (and R version 2.9.2, 2009-08-24).  Is there a way to
install an old package that still works?



More information about the R-sig-mixed-models mailing list