[R-sig-Geo] Can you krige with a nominal response variable

Edzer J. Pebesma e.pebesma at geo.uu.nl
Thu Dec 7 11:49:36 CET 2006


The indicator kriging idea does not stop at a single binary variable. 
The "standard" (or "Stanford"?) way of getting non-negative predictions, 
and possibly predictions that sum to one is rather pragmatic--simply 
postprocessing predictions. There is a cleaner solution, called 
compositional kriging, which adds the "predictions should be 
non-negative and sum to one" constraint to the kriging system, but I 
don't know if implemenations are available in R. For compositional data 
analysis you may want to explore the R package compositions. I believe 
it does not do compositional kriging, but perhaps its author has 
suggestions; I Cc:'d this email to the author.

Best regards,

Pierce, Ken wrote:
> Is there a method for kriging a response with multiple states, for
> instance vegetation class in 10 classes? I know indicator kriging would
> work for a boolean variable. Is there an extension for multiple
> responses?
> Kenneth B. Pierce Jr.
> Research Ecologist
> Landscape Ecology, Modeling, Mapping and Analysis Team 
> PNW Research Station - USDA-FS 
> 3200 SW Jefferson Way,  Corvallis,  OR 97331 
> ken.pierce at oregonstate.edu
> 541 750-7393 
> http://www.fsl.orst.edu/lemma/gnnfire
> _______________________________________________
> R-sig-Geo mailing list
> R-sig-Geo at stat.math.ethz.ch
> https://stat.ethz.ch/mailman/listinfo/r-sig-geo

More information about the R-sig-Geo mailing list