R help - Jan 2003 - link function for gaussian family

Dear All,

Is it possible to apply "log" link to gaussian family in glm?

Help file saids that the gaussian family has only one permissible link,
 "identity".  Similar descriptions are also in "Modern Applied
Statistics with
S-plus".  But the application of log link to gaussian family seems to be
valid
in R.

Thanks for any help.

Sincerely,

Kazuki Miyamoto

*****************************************
Kazuki Miyamoto (Ph. D.)

Kansai Research Center, Forestry and Forest
Products Research Institute,
Nagaikyutaro 68, Momoyama, Kyoto 612-0855,
Japan

Tel: +81.75.611.1385
Fax: +81.75.611.1207
E-mail: mkazuki at ffpri.affrc.go.jp
*****************************************

On Fri, 24 Jan 2003, Kazuki Miyamoto wrote:
> Dear All,
> 
> Is it possible to apply "log" link to gaussian family in glm?
> 
> Help file saids that the gaussian family has only one permissible link,
>  "identity".  Similar descriptions are also in "Modern
Applied Statistics with
> S-plus". 
Well, no choice of link is allowed in S-PLUS (sic), the subject of the 
book you mention.

The R help file for gaussian() is wrong, and I will correct it.
"inverse",
"log" and "identity" links are allowed in the R code.  This
has been so
since the current  CVS file started in March 1998.
> But the application of log link to gaussian family seems to be valid
> in R.
Note that it is quite possible for no MLE to exist in that case, so use 
with care: the symptom will be convergence problems, but they can occur 
even if there is a valid solution.

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595