I have found anomalous behavior in gstat's variogram estimation. I have listed
3 example variograms below for small data sets. In order to better estimate
the nugget effect, I slightly perturbed the locations (by 1 foot increments) of
duplicate results. The empirical variograms are given
-
Van: r-sig-geo-boun...@stat.math.ethz.ch
[mailto:r-sig-geo-boun...@stat.math.ethz.ch] Namens Carson, John
Verzonden: maandag 12 oktober 2009 15:35
Aan: r-sig-geo@stat.math.ethz.ch
Onderwerp: [R-sig-Geo] Problem with gstat variogram estimation
I have found anomalous behavior in gstat's variogram
John, thanks for sharing this with r-sig-geo.
As Thierry mentioned, the default model fitting procedure (fit.variogram
in package gstat) uses weighted least squares, with weights proportional
to N_h/(h^2). This explains why the first lag gets so much weight.
For pure nugget models, this of
Edzer Pebesma wrote:
John, thanks for sharing this with r-sig-geo.
As Thierry mentioned, the default model fitting procedure (fit.variogram
in package gstat) uses weighted least squares, with weights proportional
to N_h/(h^2). This explains why the first lag gets so much weight.
For pure
: ONKELINX, Thierry [mailto:thierry.onkel...@inbo.be]
Sent: Monday, October 12, 2009 10:23 AM
To: Carson, John; r-sig-geo@stat.math.ethz.ch
Subject: RE: [R-sig-Geo] Problem with gstat variogram estimation
Dear John,
Isn't your problem rather a result from the unstable empirical variograms?
Mainly due
://www.shawgrp.com/
Shaw(tm) a world of Solutions(tm)
-Original Message-
From: Edzer Pebesma [mailto:edzer.pebe...@uni-muenster.de]
Sent: Monday, October 12, 2009 10:38 AM
To: Carson, John
Cc: r-sig-geo@stat.math.ethz.ch
Subject: Re: [R-sig-Geo] Problem with gstat variogram estimation
John