[R-sig-Geo] interpolating presence absence data with external drift

[Klaus Ecker]

Hello

I want to incorporate an external trend when 
interpolating binary data (presence absence data 
from vegetation records). To my understanding 
universal kriging does not work with binary 
response data and does not apply generalized 
linear models for the covariates. Does geoRglm provide these opportunities?

Kriging residuals from a deterministic model 
(regression kriging) would allow for logistic 
regression models to be used for the trend. 
However, applications of regression kriging in 
ecology are rare, especially in recent years. Is there anything wrong about it?

With logistic regression the variogramm is 
commonly calculated on standardised (Pearson) 
residuals. But I have never seen how to back 
transform them before they are added to the 
logistic trend model. Is it correct to multiply 
the kriging predictions by the square root of the 
variance of the response residuals?

Many thanks

Klaus





______________________________________________


Klaus Ecker
Swiss Federal Research Institute WSL
Biodiversity and Conservation Biology
Spatial Ecology
Zürcherstrasse 111
CH-8903 Birmensdorf (SWITZERLAND)

phone: +41-44-7392-378
fax: +41-44-7392-215
e-mail: klaus.ecker at wsl.ch
internet: http://www.wsl.ch