[6469] in s-news-athena
Spatially autocorrelated counts
daemon@ATHENA.MIT.EDU (Ottar N. Bjornstad)
Tue Feb 21 10:35:00 1995
Date: Tue, 21 Feb 1995 09:29:31 +0100
To: s-news@utstat.toronto.edu
From: bjornsto@ERE.UMontreal.CA (Ottar N. Bjornstad)
Dear all,
I wonder if anyone can give me a hand with the following problem:
I have a data set of counts of catches of mice, which I want to relate to
some envrinmental variables. The natural choice of model is to use
something with poisson error. The log-link seem to do reasonably well.
However, my animals are territorial so that the dat set contains a fair
amount of positive autocorrelation (roughly at the scale of the size of the
territory). This is easily seen in the empirical variogram of the residuals
of any fitted model.
Had the data been Gaussian and the modelling framework had been linear
models, this problem could have been approached by using some type of
generalised least-squares. Perhaps something like the variogram-GLS method
(e.g., Cressies 1991).
However, my data are not gaussian and I use GLM models. Does anybody have
some advice on how to make a model for this situation in S-plus? Is there
anyway of using a variance-covariance matrix for poisson-type of models?
Any help would be appreciated.
Sincerely,
Ottar Bjornstad
--------------------------------------------
Ottar N. Bjornstad
Until 10th of may 1995:
bjornsto@ere.umontreal.ca
Univ. Montreal., Dept. Sci. Biol., CP 6128, Montreal, H3C 3J7, Canada.
Tel. (514) 343 6111 ext 1233, Fax. (514) 343-2293
After:
ottarnb@bio.uio.no
Div. Zool., Dept. Biol., Univ. of Oslo, Box 1050 Blindern, 0316 Oslo, Norway
Tel. + 47 22857295, Fax. + 47 22854605
