> Daniele Iannuzzo wrote:
>
> - GSTAT by default uses pseudo-cross-variograms to analyze the spatial
> continuity between different variables. Is there a way to use even
> classic cross-variograms, to test if are there important differences
> between the two algorithms?
Yes. Make sure your data file contains only those points that have
measurements of both variables at each point location. Then, the
classical cross variogram becomes the default.
>
> - We had problems with 'nugget effect', because often it was so strong
> to produce trivial maps with spikes at sample locations and values
> very close to sample mean elsewhere. In 'Multivariate Geostatistics'
> by Wackernagel, it was suggested to delete the value of the Nugget
> only in the right hand side matrix of the kriging system. Is there a
> simple way to do this with GSTAT?
Yes. Use Err() instead of Nug() to denote measurement error instead
of nugget variance (which refers to micro variability). Look in the
manual for a longer discussion on this (`Kriging data with known
measurement errors); I also recall that Cressie's
book has an excellent discussion on this point.
>
> - Is there a command to use cokriging with only one unbias condition,
> instead of classical cokriging with the sum of weights of correlates =
> 0 ?
Yes. Given:
data(a): 'file', ...;
data(b): 'file', ...;
adding the command:
merge a with b;
will force one unbiasedness condition (sum of all weights equals one)
instead of the default two (sum of weights for primary variable equals
one, for secondary variable equals zero).
Best regards,
--
Edzer
