Multivariate lme

[Sven.Knudsen@adeptscience.dk]

Dear S group

I would like to fit a multivariate variance component
model. 

The variance component part by using pdBlocked and
pdIden seem to be a viable aproach, but how can the multivariate dependence
be included?

To be specific, let's say the 1-dimensional model
is a classic 2-way variance component analysis, something like

>lme(pre.mean~1, random=pdBlocked(list(pdIdent(~maskdim
-1),pdIdent(~visc.tem-1))),data="">

Any suggestions how to analyze correlated y1 and y2
in multivariate setting are valued. In manova it would look like

>manova(cbind(pre.mean, post.mean)
~ maskdim + visc.tem, wafer) 

Thanks

Regards
Sven