Skip to content

xvalid or similar with krige.bayes()

3 messages · Damian Maddalena, Paulo Justiniano Ribeiro Jr

#
Greetings:

I need to perform LOOCV for krige.bayes().  I want to compare the output 
of krige.bayes to the output of other interpolation methods. I can do 
LOOCV easily for other interpolation methods using xvalid() in geoR or 
krige.cv() in gstat but I cannot find any code examples that demonstrate 
how this can be done for a krige.bayes() object.

Any help would be greatly appreciated.

Thank you.

-Damian
#
Well this is not that straightfoward (or uniquely defined)
since you need to specify what would 
be the "point prediction" in the Bayesian kriging.
The posterior mean is a candidate, but not uniquely defined.
Besides the analysis will be quite timing consuming
when running krige.bayes() for each of the location left aside

I personally would go for other measures of fit
but if you realy find CV within the Bayesian approach you can write
a simple loop call to the leaving-one-out



Paulo Justiniano Ribeiro Jr
LEG (Laboratorio de Estatistica e Geoinformacao)
Universidade Federal do Parana
Caixa Postal 19.081
CEP 81.531-990
Curitiba, PR  -  Brasil
Tel: (+55) 41 3361 3573
VOIP: (+55) (41) (3361 3600) 1053 1066
Fax: (+55) 41 3361 3141
e-mail: paulojus AT  ufpr  br
http://www.leg.ufpr.br/~paulojus
On Tue, 12 Mar 2013, Damian Maddalena wrote:

            
#
Thank you for your reply, Paulo.

As long I can find a statistically valid method to compare the Bayesian 
kriging surfaces to the others I will be happy. I don't necessarily need 
CV. It was just what I knew for the other interpolation methods.

Please forgive my simple question, but can you please suggest a few 
terms that I might use to search for "others measures of fit?" Something 
that I could search to find some good code examples.  I am fairly new to 
coding in R and having something to examine and modify would be a big help.

Thank you.

-Damian
On 03/13/2013 01:37 PM, Paulo Justiniano Ribeiro Jr wrote: