d. 06/01/01 21:33 skrev Berterretche, Mercedes på ... Cokriging should prove better if the secondary variable provides useful information about the primaryMessage 1 of 2 , Jan 10, 2001View Sourced. 06/01/01 21:33 skrev Berterretche, Mercedes på
> I'm confused because crossvalidation MSE (taking one sample out andCokriging should prove better if the secondary variable provides useful
> recalculating) and validation MSE (taking 15 percent of the samples out and
> recalculating) are giving me opposite results. The validation method would
> allows me to compare kriging vs cokriging vs Kriging with an external drift
> vs regression , but I don't know if I can trust the results at this point.
information about the primary one. In the worst case it should theoretically
be equivalent to kriging if the primary and secondary variable are
uncorrelated. Kriging with an external drift using a drift function that's
totally out of whack and unrelated to the primary variable can prove
to be worse than normal kriging. Likewise it can be better than kriging if
the drift provides useful information on how the shape (or trend) of the
variable is distributed in space. "Regression" is kriging assuming no
spatial correlation. Cokriging should always be better than or equivalent to
kriging which likewise should be better than or equivalent to "regression".
Better, though, is always a subjective word. KED is always "better" than
cokriging because it eliminates tedious cross-covariance modeling. Likewise
neural networks to incorporate more than two secondary variables. MSE is
just one tip of the iceberg. Frequently a transfer function such as a
fluid flow simulator will determine whether the result is "better" or more
realistic. In such a case perhaps sequential simulation will prove "better."
What is "best" or "better" is usually quite an elusive concept, and wholly
dependent on different people's perceptions of what constitutes quality in
their minds taking into account prior and existing prejudices.
* To post a message to the list, send it to ai-geostats@...
* As a general service to the users, please remember to post a summary of any useful responses to your questions.
* To unsubscribe, send an email to majordomo@... with no subject and "unsubscribe ai-geostats" followed by "end" on the next line in the message body. DO NOT SEND Subscribe/Unsubscribe requests to the list
* Support to the list is provided at http://www.ai-geostats.org