[R-sig-Geo] gstat: Variogram with non-Euclidean Distances

[Julia Dechamps]

Ben and Luca,

Thank you very much for your advice and the references.

When using non-Euclidean distance I do precisely run into issue of obtaining non-positive definite correlation matrices,  so I might have to return to Euclidean distance...

Regards,
Julia


________________________________________
From: Julia Dechamps
Sent: 11 August 2011 10:29
To: r-sig-geo at r-project.org
Subject: gstat: Variogram with non-Euclidean Distances

Dear list,

I am wondering whether there is a way to use a contiguity-, graph- or knn-based measure of distance (similar the neighbourhood objects described in "Applied Spatial Data Analysis with R, ch.9.2.2-9.2.4), rather than Euclidean distance, to construct a variogram (in gstat or otherwise).

The observations I have are very unevenly spaced across a large geographic area and I therefore feel Euclidean distance might not be a good choice.

I have searched the web but unfortunately was not able to find any hints on how I might implement the above.

I would greatly appreciate any advice you might have.

Regards,
Julia