R help - Apr 2010 - mixed model with subject-specific variance structure

Hi All,

I built a mixed model with subject-specific variance structure in SAS and it
works quite well. Now I intended to implement the same model in R using lme with
"weights" option. Following is the SAS code and the corresponding R
code:

SAS:
proc mixed data=source;
class id trial group valence scan;
model y=group|valence|scan;
repeated /sub=id(trial) group=id type=AR(1);
run;

R

vf1Ident<-varIdent(~1|id)
vf1Ident.ini<-Initialize(vf1Ident,grouped.data)
model<-lme(fixed=y~group*valence*scan,data=grouped.data,correlation=corAR1(),weights=vf1Ident.ini);

The result shows :convergence error code = 1  message = iteration limit reached
without convergence (9)

I searched the help archive and likely the weights option does not work well.
Any suggestion to implement the subject-specific variance structure in R?

Thanks,

Yonghua

	[[alternative HTML version deleted]]