[R] Think I'm sure, but confirm: lme4 vs. nlme

Bert Gunter gunter.berton at gene.com
Wed Oct 22 17:23:10 CEST 2008

Doug Bates is the authority of course, but perhaps a couple of quick inline
comments may suffice:

-----Original Message-----
From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org] On
Behalf Of Mitchell Maltenfort
Sent: Wednesday, October 22, 2008 7:41 AM
To: r-help at r-project.org
Subject: [R] Think I'm sure, but confirm: lme4 vs. nlme

The impression I get from the list and the references I've perused is
that nlme is being phased out in favor of lme4, but lme4 still doesn't
have a complete feature set yet.

-- Not exactly. nlme isn't really being phased out (yet anyway) -- it's just
not being improved anymore. However, it is a pretty nice implementation of
mixed effects models for continuous data with Gaussian errors with nested
random effects. And not too large data sets. I would opine that it handles
the bulk of mixed effects models that one encounters in the physical and
biological sciences very well.

What I'm still fuzzy on, being a relative R newbie, is:

(a) what features in nlme are currently missing in lme4
(b) what's the projected time frame on getting them implemented.

-- I think the time frame is "continuing, as Doug is able to work on it." As
a user, I would say that this is not a discrete event, but an evolving work.
I use lme4 routinely now.
As for features, that is difficult to answer. Last I checked, nlme has a
good predict method but lme4 didn't. lme4 does mcmc for the posterior
distribution of parameters in Gaussian models, nlme doesn't; but nlme does
an approximate version of confidence intervals for parameters. Also lme4
handles glm's, crossed random factors, very large data sets; nlme does none
of this. Etc. Doug may, of course, have corrections or additions here.

If anyone can answer my naive and impertinent question in a friendly,
helpful and informative manner[1], I would be delighted.  Thanks!

-- I didn't consider your questions naive or impertiment. I leave it to you
to judge whether I have been friendly or helpful.

Bert Gunter

[1] Or, translated from craven, "please don't ignite a flame war on my
no-account account"


Due to the recession, requests for instant gratification will be
deferred until arrears in scheduled gratification have been satisfied.

R-help at r-project.org mailing list
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.

More information about the R-help mailing list