[R] GLM with a multinomial response

[Petr Smilauer]

Dear all,
  does anyone know how to setup a generalized linear model for data looking
like
the following example. This would be an experiment on food preferences for,
say,
snails, and I would measure for each specimen its choice made on food type
food
plus one or more quantitative (or factorial) explanatory variables, here
represented
by the snail size.
SpeciesSelected    SnailSize
    Spc1                      2.7
    Spc2                      3.5
    Spc1                      2.9
    Spc3                      4.6

.....

I believe that even in the situation where my explanatory variables were
categorial,
I shall not use the loglinear models (i.e. modelling counts and moving the
"SpeciesSelected"
descriptor to the right side, studying its interaction terms with the
predictors), because
the total count has an upper bound determined by the experimental design and I
shall
model the probability of selecting particular plant species given snail size.
Doing separate models with binomial distribution for each food type is not
correct, either,
because the constrain of probabilities summing up to 1.0 is violated.

Does anyone have any thoughts on this issue? This shall not be so rare type
of analysis, given the various customer preference studies. Of course, I can
use
classification trees (or neural-network models), but I would like to use more
parametric
description of dependence.

Thank you for any help you might provide!


Best wishes
                        from

--------------------------------------------------------------
Petr Smilauer
Faculty of Biological Sciences
Ceske Budejovice, Czech Republic
--------------------------------------------------------------
Course Multivariate Analysis of Ecological Data <http://regent.bf.jcu.cz>


-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
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
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._