Sorry it have took long time to post the summary
concerning my question.
The first question,
> We read some books on geostatistics.
> In application to actual data in most books,
> it is not shown that why the semivariograms appeared in the analysis
> are selected.
> Does anyone have reference or comments ?
We have found following three criteria in some papers.
1. goodness of fit in variogram model fitting.
2. cross validation in kriging.
3. Akaike's Information Criterion with MLE.
In our analysis, criterion 1 has selected Gaussian model.
But the kriging surface with the fitted Gaussian model
has very strange feature.
The next question,
> I have one more question.
> If we assume a process is Gaussian,
> maximum likelihood (ML) estimator or
> restricted maximum likelihood (REML) estimator can be made
> for estimation of its covariance structure.
> I want to know another distributional assumptions on
> stochastic process which are often used in practice.
In analysis with density of metal or something,
many data behave like log normal distribution.
So, in this case, the data are to be log transformed.
Another transformation for non-normal data is
to disjunctive kriging.
Thank you those who have responsed my question.
Dept. Math. Sci, Fac. Eng. Sci,
Osaka Univ., Toyonaka, Osaka 560, JAPAN
*To post a message to the list, send it to ai-geostats@...
*As a general service to list users, please remember to post a summary
of any useful responses to your questions.
*To unsubscribe, send email to majordomo@...
with no subject and
"unsubscribe ai-geostats" in the message body.
DO NOT SEND Subscribe/Unsubscribe requests to the list!