Skip to content
Prev 4258 / 29559 Next

Doubts about kriging

On Thursday 02 October 2008, GEMA FERNANDEZ-AVILES CALDERON wrote:
Hi,

Dodging your questions a bit, but I have some concerns and/or issues for your 
consideration:

1. 25 locations is very small for kriging, unless you have some other means to 
develop your spatial covariance models. Perhaps you mean that the 6 variables 
are collocated at only 25 locations, but you have other sites with fewer than 
all 6? If not - if all you have are 25 sites with six variables each - I 
wouldn't attempt kriging in any form.

2. Actually I think 25 is pretty small for the multivariate analysis you 
propose as well. 

3. With respect to your point 2, I would develop the linear combination with 
the real data, not on a bunch of interpolated 'data'. Standard methods of 
combining the variables will assume that the observations are independent; if 
they aren't then the model will be flawed, so be cautious of that.

Hope this helps,

Ashton