R help - Oct 2005 - syntax for interactions in lme

Hello,

I am trying to make the switch from SAS, and I have a fairly elemental 
problem with syntax using the nlme package for analyzing mixed models.  
There was a previous question on this topic posted to this list, so I 
apologize for redundancy, but I didn't understand the advice given to 
that inquiry.  The model I want to run has the following factors:

Host (fixed)
Sire (random)
Dam nested within Sire (random)
Host * Sire (random)
Host * Dam within Sire (random)

So without the interactions I have:

hogmodel = lme(gain ~ host, random = ~1|sire/dam)

If I understand correctly, that "sire/dam" term gives me both Sire and
Dam within Sire as random factors.  OK, so now I want to add the two 
interactions (listed above)...

In SAS, I would just enter them as random factors, but I can't seem to 
make that work.  I think I need to make a list that includes Sire/Dam, 
Host:Sire, and Host:Sire/Dam.  Two questions:

1) Is it sufficient to have simply Host:Sire/Dam in order to code for 
both interactions (host*sire and host*dam-within-sire)?

2) In any event, WHAT IS THE SYNTAX FOR MAKING A LIST OF RANDOM FACTORS?

I have looked in section 1.3 of Pinheiro&Bates, which was helpful, but 
I wasn't able to answer my question.  Thanks in advance, your help will 
keep me from running back to SAS!

--Matt



- - -
Matthew L Forister
Department of Ecology and Evolution
State University of New York at Stony Brook
650 Life Sciences Building
Stony Brook, New York 11794-5245
Email: forister at life.bio.sunysb.edu
Webpage: http://life.bio.sunysb.edu/~forister/
Lab phone: (631) 632-8609
Fax: (631) 632-7626
- - -

> Host (fixed)
> Sire (random)
> Dam nested within Sire (random)
> Host * Sire (random)
> Host * Dam within Sire (random)
> 
> So without the interactions I have:
> 
> hogmodel = lme(gain ~ host, random = ~1|sire/dam)
> 
> If I understand correctly, that "sire/dam" term gives me both 
> Sire and Dam within Sire as random factors.  OK, so now I want
> to add the two interactions (listed above)...
Correct, for the interactions write:

random = ~ host | sire/dam

an interaction between a fixed and a random term can be interpreted as a
variability in the fixed term for the different levels in the random term.
Thus the 1 (which stands for the intercept) is exchanged with the fixed
effect(s) with which interactions are of interest.

This is easy if all the hierarchical levels of a nested random effects go
into an interaction it is a bit more complicated if not. Say you only want
the interaction of host and dam but not sire:

random = list (~ 1 | sire, ~ host | dam)

Hope this helps (it can all be found in Pinheiro and Bates and on the help
pages).

Lorenz