[R] Stratifying level-1 variance with lmer()

[Dieter Menne]

David Afshartous <dafshartous <at> med.miami.edu> writes:

> I've fit some models via lme() and now I'm trying to fit similar models with
> lmer() for some simulations I'm running.
> 
> The model below (fm1) has an intercept variance that depends on treatment
> group.  How would one accomplish a similar stratification for the level-1
> variance, i.e., the within-group or patient variance?
> 
> With lme() I was able to do this via:
> update(fm1,
> weights = varIdent(form = ~ 1 | treatment.ind) )
> 
> However, this does not work for lmer().  Is there any way around this?

This is not yet implemented. See Douglas Bates'

http://article.gmane.org/gmane.comp.lang.r.lme4.devel/518

Dieter