[R] Nested Design Coding Question

[Steeno, Gregory S]

I'm a SAS user who is slowly but surely migrating over to R.  I'm trying to
find the proper code to analyze a nested design.  I  have four
classification variables, L (fixed), A (random within L), D (random within
L), and I (random within L).  The model I'm interested in is

L A(L) D(L) I(L) A:D:I(L),

where the interaction is interpreted as the lack-of-fit term.  I've tried
variants of the lme function similar to these,

lme(response~L, data, random=~Lab/(A+L+I+A:D:I),
lme(response~1, data, random=~Lab/(A+L+I+A:D:I),
lme(response~L, data, random=~1/(A+L+I+A:D:I).

All give results different from SAS, and all give warning messages regarding
either false- or non-convergence.

For reference, the abbreviated SAS code is,

model response = L;
random A(L) D(L) I(L) A:D:I(L);

Can anyone shed some light?  I'd be very appreciative.

Thanks.

Greg Steeno









LEGAL NOTICE\ Unless expressly stated otherwise, this message is ... [[dropped]]