Page 1 of 1

neighborhood and cokriging

PostPosted: Mon Jun 15, 2015 12:29 pm
by Maxime Beauchamp
Hi,

I am wondering if it is possible to define different neighborhoods in RGeostats when I carry out a cokriging. In fact, I'd like to use a large dmax for my first variable Z1 and smaller one for my second variable Z2. Is it allowed, by setting a vector of 2 values in the dmax arguments of the neigh.init function ?

Yours sincerely,

Maxime Beauchamp

Re: neighborhood and cokriging

PostPosted: Tue Jun 16, 2015 7:12 am
by Didier Renard
Sorry. This feature does not exist.

The reason is that I am afraid that all the parameters could be important (not only dmax). In other words, the neighborhood for cokriging could be a multiple-neighborhood (with say as many neighborhood as there are variables). This is simply not tractable as one of the main question afterwards is to determine a policy for selecting samples: i.e. a sample where two variables are defined is by far more important than two samples with one variable each.

The solution which has been implemented is the collocated neighborhood which is automatically initiated as soon as the collocated cokriging is launched.
The principle is:
- we consider that there is a main variable and one or several auxiliary varibales
- select a neighborhood based on the main variable and select 'Nmain' samples
- any variable also defined on 'Nmain' samples will be considered
- as much variables as possible are considered at the target location

Hope this will help.

Re: neighborhood and cokriging

PostPosted: Tue Sep 22, 2015 7:22 am
by MaximeBeauchamp
Hi,

Thanks for the answer. I will then focus on the collocated cokriging if the problem comes again.