[R-sig-ME] weighted-ML estimates

[Douglas Bates]

On Fri, Feb 13, 2009 at 8:55 AM, H c <harlancampbell at gmail.com> wrote:
> Hi,
> We are trying to calculate ML-parameter estimates of a mixed effects models
> where the observations are weighted.  The "weights" option in lmer() with
> RELM=FALSE seems attractive.  Does anyone know the mechanism it uses to
> calculate weighted ML estimates?

The weights are with respect to the observations.  The log-likelihood
depends on a residual sum of squares in the unweighted case.  In the
weighted case a weighted residual sum of squares is used.

> (Is there a paper?).  Also, our model
> includes a serial correlation structure(e.g. AR(1)) among the residuals.  As
> far as I know, no such "cor" option is available in lmer() unlike in nlme.
>  Is this correct and if so, any suggestions on how to proceed?

You are correct.  The lmer function does not incorporate serial
correlation structures.