[R] model.tables

John Maindonald john.maindonald at anu.edu.au
Wed Dec 1 04:21:11 CET 1999

Berwin Turlach wrote -

> > >>>>> "JM" == John Maindonald <john.maindonald at anu.edu.au> writes:
>   JM> At 08:02 30/11/99 +0000, Prof Brian D Ripley wrote:
>   >> On Tue, 30 Nov 1999, spoon <spoon at hilbert.maths.utas.edu.au> wrote:
>   >> 
>   >>> Hi,
>   >>> Is this a bug or do I just not understand model.tables?
>   >>> [...]
>   >>> Or am I just completely misinterpreting something basic?
>   >> 
>   >> Basically, yes. This is an incompletely replicated design, and d and e
>   >> occur on different litters.
>   >> 
>   >> The results are identical to the S-PLUS original. I think you are
>   >> probably looking for what dummy.coef gives you.
>   >> [...]
>   JM> So what is it that model.tables() gives?
> Good question, in R (0.65.1) the documentation of model.tables says:
>         The implementation is incomplete, and only the simpler
>         cases have been tested thoroughly.

Simon and I had noted that.  As S-PLUS has no such warning, perhaps
the discussion should be taking place on s-news.

>   JM> S-PLUS says they are estimates.  Of what?  These are not the
>   JM> marginal means of the fitted values, ignoring other factors.  Do
>   JM> they mean anything at all?  The issue of unbalance does not
>   JM> arise here.
> The documentation in R also states that
>     Details:
>         For `type = "effects"' give tables of the coefficients
>         for each term, optionally with standard errors.
>         For `type = "means"' give tables of the mean response
>         for each combinations of levels of the factors in a
>         term.
> Hence, my understanding would be that if `type="means"' is specified
> the marginal means of the observations ignoring other factors is
> given.  But this doesn't seem to be the case, at least not with an
> unbalanced design.  You can use either your dreamed up example or the
> data that spoon has posted.

On the usual definitions of `balance' both in my example and
Simon's example treatments are `balanced" over blocks (or litters).


John Maindonald               email : john.maindonald at anu.edu.au        
Statistical Consulting Unit,  phone : (6249)3998        
c/o CMA, SMS,                 fax   : (6249)5549  
John Dedman Mathematical Sciences Building
Australian National University
Canberra ACT 0200
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch

More information about the R-help mailing list