[R-sig-Geo] kriging with external drift problems

[Edzer Pebesma]

Dear Els, I believe there is no cure for too little data.

Instead of looking at directional variograms, I would look at variograms
of regression residuals. Instead of a max distance value, I would set a
maximum neighbourhood size expressed as the max number of nearest values.

Whether 43 observations can be sufficient to indicate strong
non-stationarity, is questionable; cross validation might give you some
information about how helpful it is to work in local neighbourhoods, but
again, it only has 43 observation to do this.

As usual, when going from a global to a local model you expect a
decrease in bias, and an increase in variability.

On 12/09/2011 01:23 PM, Els Verfaillie wrote:
> Dear list,
> 
>  
> 
> I have a dataset from Ethiopia with a good correlation between infiltration
> depth and slope (r=0.75). Therefore, I would like to use KED (with slope as
> secondary variable) to interpolate the infiltration observations.
> 
> However, my main problem is the limited amount of observations (n=43). This
> gives problems for the directional variogram calculation and for the use of
> an interpolation window. I would like to consider the direction where the
> drift is not present as underlying variogram, but with the very limited
> amount of observations this is not straightforward. 
> 
>  
> 
> My questions are the following:
> 
> -          Is there an alternative interpolation technique for my approach
> given the small number of observations, if possible making use of the slope
> information as secondary variable?
> 
> -          Because of the strong non-stationarity of the dataset I want to
> use a local interpolation window. But when I set a distance value I receive
> an error message ("lufactor.c", line 207: singular matrix in function
> m_inverse()). Is there a way to solve this problem? 
> 
>  
> 
> Thank you for your answer!
> 
>  
> 
> ______________________________________________
> 
> Els Verfaillie
> 
> Ghent University (UGent) - Department of Geography
> 
> Belgium
> 
> ______________________________________________
> 
>  
> 
> 
> 	[[alternative HTML version deleted]]
> 
> _______________________________________________
> R-sig-Geo mailing list
> R-sig-Geo at r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-geo

-- 
Edzer Pebesma
Institute for Geoinformatics (ifgi), University of Münster
Weseler Straße 253, 48151 Münster, Germany. Phone: +49 251
8333081, Fax: +49 251 8339763  http://ifgi.uni-muenster.de
http://www.52north.org/geostatistics      e.pebesma at wwu.de