[R] random effects structure in lme

Vito Muggeo vito.muggeo at giustizia.it
Tue Apr 20 16:09:01 CEST 2004


Dear all,
In using the lme() function from the nlme package, I would like to specify a
particular correlation structure for the random effects.
For instance, for a 3 by 3 matrix, say, I am interested in assuming a
`quasi-diagonal'
matrix having zero entries in all but the first column (and the main
diagonal, of course)

Hence, given the three random effects b1 b2 and b3, I want to assume
cov(b2,b3)=0;
given four random effects b1,b2,b3,b4 it should be
cov(b2,b3)=cov(b4,b2)=cov(b4,b3)=0.

For a simple diagonal matrix I specify `random=list(id=pdDiag(~1+x1+x2))',
(id is the grouping variable), but I am not able to go on....

Please, could anyone give me any advice?

many thanks,
vito




More information about the R-help mailing list