[R-sig-Geo] pcnm - scaling

robert.ptacnik at niva.no robert.ptacnik at niva.no
Wed Dec 17 18:51:09 CET 2008



Dear listers,

[below message was posted on R-help already, but maybe is pøaced better
here]

I am using the pcnm function (spacemakeR) to obtain eigenvectors for a

spatial grid of sampling sites. These pcnm eigenvectors are then used in

multivariate ordination to test where community composition follows local

environment or rather shows spatial autocorrelation, and get support for

apatial autocorrelation.

In an RDA analysis, I can see which eigenvectors correlate most with my

data, which can be interpreted that the spatial scale is rather 'fine' or

'broad' (broad in case of my data). However, is there a way to read the
spatial scale on which the

data correlates best, i.e. is it possible to extract a spatial scale

describing the spatial autocorrelation (e.g. from the eigenvalues)?

My data


- irregularly spaced sampling sites, total scale ca. 2000 km, most sites

few tens of km apart

- i use the dfunction rdist.earth to obtain a distance-matrix of all sites

prior to using pcnm

Thanks !

Robert



More information about the R-sig-Geo mailing list