[R] Weights in mixed models

[David R.]

Hello everyone,

I wonder if sample size can be used as weight  in a weighted mixed model. Or 
should I use just the inverse of the variance?

For example, the class'lmer' in the 'lme4' package have an option 'weights'
(as in the class 'lm' of 'stats'). In the help of lme4 there is an example using 
'herd size' as weight in a mixed model.
 
 So, if I have a measurement data (eg height) of 10 groups (sample size ranging 
from 30 to 3000 individuals for each group) can I  use this number (N, sample 
size) in the 'weights' option? Is this wrong? 

 
Finally, what to do if the results (coefficients) of weighing by 'inverse of the 
variance' or by 'sample size' are very different, even opposite?


Thank you very much in advance
David