Skip to content

autologistic model in R

3 messages · Nicola Batchelor, Paulo Justiniano Ribeiro Jr, Kamran Safi

#
Dear all,

 

I'm currently using a logistic regression model for my dataset, which has a
binary outcome of 0 or 1.  I've looked at the residuals, which are spatially
autocorrelated, so I want to do an autologistic regression model to remove
any residual correlation.

 

I've been looking online, and it seems there are not yet any packages with
such a function.

 

I've found references to Rcitrus which seems like it should be of use, but
the package doesn't seem to be available at the moment for some reason.  I
have also seen reference to the stLattice package which is under
construction.

 

I'm a novice to this type of analysis, so I'd really appreciate any input as
to which would be the most suitable package, or any other advice.

 

I have considered using geoRglm, but am a reluctant as I tried it previously
and found it a bit too involved for a beginner such as myself!

 

Thanks,

Nicola

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://stat.ethz.ch/pipermail/r-sig-geo/attachments/20090114/93ebe0e1/attachment.html>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: not available
URL: <https://stat.ethz.ch/pipermail/r-sig-geo/attachments/20090114/93ebe0e1/attachment.pl>
#
Nicola

Rcitrus and stLattice are "unofficial" packages
mantained by Elias Krainski and available
from links at www.leg.ufpr.br.

Rcitrus implements a simple autologistic model motivated by the
need to analyse date from citrus fields using such model.
It also implements some other methods commonly used for such analysis such
as quadrat counts, Taylor Law, some simulation algorithms.
The documentation is only available in  Portuguese and,
although available, it is unlike to be turn into an official R package.

stLattice is meant to generalise the function to fit autologistic
model via the key fnction aglm(),
turning it into a more flexible and general setup for such models,
including possibilities to specify time lags and "auto" and "non-auto"
spatial covariates via a lag() function to be used in the formulae.

This is working in progress but some functionality is alread available and
the Session " Paper Companions" at the LEG web page has an example code
used in a paper.

Hope this helps and you can contact us via the email available at the page
for further details

Best
P.J.


Paulo Justiniano Ribeiro Jr
LEG (Laboratorio de Estatistica e Geoinformacao)
Universidade Federal do Parana
Caixa Postal 19.081
CEP 81.531-990
Curitiba, PR  -  Brasil
Tel: (+55) 41 3361 3573
Fax: (+55) 41 3361 3141
e-mail: paulojus AT  ufpr  br
http://www.leg.ufpr.br/~paulojus
On Wed, 14 Jan 2009, Nicola Batchelor wrote: