[R-sig-Geo] comparing spatial distribution of two species in a same area with different coordinates

Virgilio Gómez-Rubio virgilio.gomez at uclm.es
Mon Apr 18 14:44:34 CEST 2011


El lun, 18-04-2011 a las 04:58 -0700, Corey Sparks escribió:
> Would the cross L or K function not give you a descriptive tool for what you
> want? they measure clustering in two or more point patterns. See 
> ?Lcross
> or 
> ?Kcross
> in spatstat.  Also there are inhomogeneous Poisson process models as well,
> Lcross.inhom

As I have understood this problem, there are aggregate data at different
locations. So the point pattern approach does not seem feasible.

A possible alternative could be using a geostatistical model with a
Poisson response, so that you have a spatially correlated effect
underneath. You may fit the same model to your data sets and then
compare the  spatially correlated effects.

You could use geoRglm. In particular, functions pois.krige and
pois.krige.bayes seem to be relevant. And you can also check the book on
model-based geostatistics:


Could other people comment on this approach?

I can think of other interesting (Bayesian) models. If you are happy to
go this way, feel free to contact me off-list.



More information about the R-sig-Geo mailing list