> Is there a possiblity to estimate classification and regression tree
> models taking spatial dependence into account?
> Or are there any viable functions in R to model spatial data with
> ordinal response, e.g. using something like autoregressive error
> Thanks for your advice!
Scott Urquhart at Colorado State University (nsu at stat.colostate.edu)
developed with a student a pre-whitening approach to handling spatial
dependence in regression trees. There are a set of R/Splus functions
that I believe are available. If you contact him and decide that you
want to use these, and he doesn't have them, I think I can find them.