[R-sig-ME] plotting random effects; lme4(a) inference methods

Robert Kushler kushler at oakland.edu
Sun Apr 4 18:38:18 CEST 2010

Page 70 of "Ch4.pdf" (the draft of the new book) says:

"The plot method for "ranef.mer" objects produces one plot
for each grouping factor."

Unfortunately the plots for all but the last grouping factor
are instantly replaced in the plot window.  My attempts to
work around this have so far failed.  Am I missing something
obvious?  (Wouldn't be the first time.) Any help would be

As long as I am posting, I will express my hope that Doug Bates
will weigh in on the recent post by Iasonas (Jason) Lamprianou, and
indicate where lme4a is heading.  The new book indicates that Doug
has (re)turned to favoring likelihood based inference methods, but
it would be nice to know the status of the Bayesian approach as well.

The latest version of mcmcsamp in lme4 seems much less prone to the
"stuck at zero" issue that occurred in earlier versions, but in my
experience (with a number of "toy problems" from linear models texts)
the estimates produced by lmer are near, or even above, the upper end
of the HPD intervals, which leads to a certain lack of "confidence"
(sic) in those intervals.

Perhaps this is a consequence of the prior that is used, but that
just raises another question:  will a future version allow user
control of the specification of the prior?

Regards,   Rob Kushler

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