Dear S-plus users,
I'm trying to fit a nested mixed effect model, with stratified
correlations on the lower level. This is the stucture of my data:
9 matched pairs with one subject each from Group 1 and Group 2
2 observations per subject corresponding to two locations
I want to fit a model with subject effect nested into pair effect, with
observations taken on the same subject being correlated, with correlation
r1 for subjects in Group1 and r2 for subjects in Group2.
The following nested random effect declaration:
random=list(pair=pdIdent(~1),subjectid=pdSymm(~location-1))
works but fits the same correlation r=r1=r2.
When I try
random=list(pair=pdIdent(~1),subject=pdSymm(~location:group-1)))
it complains, and it's absolutely right, that there are too
many random effects (one subject has observations to one group only).
So how do I make it fit the model with unequal correlations?
Hanga Galfalvy
|