[R] analysing non-normal spatially autocorrelated data

[Ruben Roa]

> -----Original Message-----
> From: r-help-bounces at stat.math.ethz.ch
> [mailto:r-help-bounces at stat.math.ethz.ch]On Behalf Of Carsten Dormann
> Sent: 20 July 2005 05:41
> To: r-help at stat.math.ethz.ch
> Subject: [R] analysing non-normal spatially autocorrelated data
> 
> 
> Dear fellow R-users,
> 
> I wish to analyse a lattice of presence-absence data which 
> are spatially autocorrelated.
> 
> For normally distributed errors I used gls {nlme} with the 
> "appropriate" corStruct-method.
> Is there any method for other families (binomial and poisson)?
> 
> A method that look suitable to me as a non-statistician is 
> called gllamm  (generalised linear latent mixed model), by Rabe-Hesketh et 
> al (2001), available apparently only for Stata.
> In R, I found the gamm {Matrix} function doing what I want, but I am 
> interested in the parameter values of the covariates, using the model 
> for prediction, hence gamm is no option.
> Finally, Dan Bebber posted a similar question to the R-help list in 
> September 2004 (about using corStruct in glmmPQL), but there 
> is no reply in the thread 
> (http://tolstoy.newcastle.edu.au/R/help/04/09/3103.html).
> 
> Any suggestions are highly welcome.
> 
> Many thanks,
> 
> Carsten

Check the package geoRglm, which fits by maximum likelihood a generalized 
linear mixed spatial model in the binomial or Poisson families, allowing for
covariates. geoRglm will also need the package geoR which fits the 
spatial model for continuous processes.
The authors of the packages have published several papers on theory
and applications.
Ruben