[R-sig-ME] non-nested effects in lme

[andrea.onofri at unipg.it]

Dear all,

I am trying to fit a simple model, relating to a randomised block design where both blocks (A) and treatments (B) are random effects. Coding in lmer, this model would be: 

model <- lmer(Y ~ 1 + (1|A) + (1|B))

However, I would also like to be able to 'manipulate' the correlation structure and thus I assume I have to revert to the lme function in the nlme package. In other cases I have been able to fit non-nested effects in lme by appropriately using the pdMat construct, but, after several efforts, I do not seem to succeed in this simple case. I would greatly appreciate any hints that puts me in the right direction. I thank you very much in advance.

Regards

Andrea Onofri