[R-sig-ME] complex residual variance

[Douglas Bates]

On Mon, May 31, 2010 at 8:47 AM, Sven De Maeyer <sven.demaeyer at ua.ac.be> wrote:
> Hi all,
>
> I'm not a frequent user of Lme4 and I have the habit to use mlwin for
> multilevel analyses. Nevertheless I'm willing to master lme4 further.
>
> In my research I'm often faced with the situation where level 1 variance (or
> residual variance in the mixed effects framework) isn't constant. E.g. I
> have longitudinal data on children, with multiple measurements
> (observations) of these children at a certain moment. As children get older
> these measurements seem to vary more and more. So the residual variance
> itself is a function of time. As a consequence there is heteroscedasticity.
>  Is there a possibility to explicitly model this in lme4? I know how to do
> it in mlwin, but is it possible in lmer?

Not at present.