[R] Question on structuring variances using the lme4 package

[Dieter Menne]

Catherine A. Holt <caholt <at> u.washington.edu> writes:

> I am modeling the effects of an environmental variable (X) on fish recruitment
(Y) for several,
> ecologically related species (i) using a mixed-effects model. The linear
relationship between X and Y
> includes a fixed effect that is common across all species and random effects
that vary by species. In the
> lmer() notation from the package lme4:
> 
> Model<- lmer(Y ~ X +(X|i))
> 
> Because the residuals of the model are autocorrelated and the variances differ
among species, I would like
> to include an AR(1) process, and species-specific variances. I have been
unable to do this so far in lme4.
> In the older “nlme” package, this was possible using:
> 
> Model<-update (weights=varIdent(form= ~1|i), correlation = corAR1(form = ~
years|i)
> 
> Is this possible in lme4, and if so, how? Any help would be much appreciated. 
> Cheers,

No, it is not (yet) possible. Any reason not use lme? It works, and it is
stable, and well documented (Pinheiro/Bates)? 

lmer is a package in development, it has been made public mainly to provide
feedback to the author, Douglas Bates. 

See also the special list 

http://news.gmane.org/gmane.comp.lang.r.lme4.devel

Dieter