Re: R-beta: glm bugs

Peter Dalgaard BSA (p.dalgaard@biostat.ku.dk)
17 Dec 1997 00:38:41 +0100


To: Jim Lindsey <jlindsey@luc.ac.be>
Subject: Re: R-beta: glm bugs
From: Peter Dalgaard BSA <p.dalgaard@biostat.ku.dk>
Date: 17 Dec 1997 00:38:41 +0100
In-Reply-To: Jim Lindsey's message of Tue, 16 Dec 1997 23:56:26 +0100 (MET)

Jim Lindsey <jlindsey@luc.ac.be> writes:

> Potential users of the glm function in R0.60 should be aware that a
> number of the bug fixes for glm in R0.50 (offsets, aliases, ...) that I
> sent on the list have not been implemented in R0.60. If they want to
> have (more) correct answers with glm, they should consult the archives
> of the list and make the appropriate corrections.
>   Jim

Um, the offset fixes HAVE been done, but they seem to have gotten
botched up.

This works:

summary(glm(counts ~ outcome + treatment + offset(outcome==1), 
family=poisson()))

This doesn't:

summary(glm(counts ~ outcome + treatment, family=poisson(),
offset=outcome==1))

The glaring bug is in this line in glm():

       offset <- model.offset(mf)

... which overrides the optional argument whether or not an offset()
term is present in the model formula. Easily fixed, but which offset
should take precedence if both are specified?

-- 
   O__  ---- Peter Dalgaard             Blegdamsvej 3  
  c/ /'_ --- Dept. of Biostatistics     2200 Cph. N   
 (*) \(*) -- University of Copenhagen   Denmark      Ph: (+45) 35327918
~~~~~~~~~~ - (p.dalgaard@biostat.ku.dk)             FAX: (+45) 35327907

-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
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@stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._