[R-sig-ME] [R] lme nesting/interaction advice

Andrew Robinson A.Robinson at ms.unimelb.edu.au
Mon May 12 02:05:58 CEST 2008

On Mon, May 12, 2008 at 10:34:40AM +1200, Rolf Turner wrote:
> On 12/05/2008, at 9:45 AM, Andrew Robinson wrote:
> >On Sun, May 11, 2008 at 07:52:50PM +0100, Federico Calboli wrote:
> >>
> >>The main point of my question is, having a 3 way anova (or ancova, if
> >>you prefer), with *no* nesting, 2 fixed effects and 1 random effect,
> >>why is it so boneheaded difficult to specify a bog standard fully
> >>crossed model? I'm not talking about some rarified esoteric model
> >>here, we're talking about stuff tought in a first year Biology Stats
> >>course here[1].
> >
> >That may be so, but I've never needed to use one.
> 	So what?  This is still a standard, common, garden-variety
> 	model that you will encounter in exercises in many (if not
> 	all!) textbooks on experimental design and anova.

To reply in similar vein, so what?  Why should R-core or the R
community feel it necessary to reproduce every textbook example?  How
many times have *you* used such a model in real statistical work,

> >If it's bog-standard and yet boneheaded difficult, then presumably
> >someone else would have had this problem before you.  Perhaps a search
> >of the archives will help?  If you try, you will find many qualifiers
> >to the effect that "lme isn't very well set up for crossed random
> >effects".
> 	But that avoids the question as to *why* it isn't very well
> 	set up for crossed random effects?  What's the problem?
> 	What are the issues?  The model is indeed bog-standard.
> 	It would seem not unreasonable to expect that it could be
> 	fitted in a straightforward manner, and it is irritating to
> 	find that it cannot be.  If SAS and Minitab can do it at
> 	the touch of a button, why can't R do it?

Bates has made no secret of the fact that lme was intended first and
foremost for nested designs, and that support for crossed designs is
not promised.  He has said so on many occasions, as a search would
find.  He is now working on lme4, which will support crossed designs.
It's not done yet. 

> >and you show
> >no evidence of having searched the R-help archives.  How is that
> >helpful?
> 	It doesn't seem to me to be a complaint as such.  It is a
> 	request for insight.  I too would like some insight as to
> 	what on earth is going on.  And why do you say Federico
> 	shows no evidence of having searched the archives?  One can
> 	search till one is blue in the face and come away no wiser
> 	on this issue.

At least one can know that there is an issue, which apparently
Federico previously did not.

Warm wishes

Andrew Robinson  
Department of Mathematics and Statistics            Tel: +61-3-8344-6410
University of Melbourne, VIC 3010 Australia         Fax: +61-3-8344-4599

More information about the R-sig-mixed-models mailing list