Simone,

My understanding of Intrinsic Hypothesis is that it is based on the 
stationarity of both difference (1st order) and variance of difference (2nd 
order). So the statement your wrote " Intrinsic hypothesis is different from 
second order one mainly because in the first case covariance function does not 
exist and variogram is computed instead of it...." does not make sense to me.


>The problem is how to realize about the intrinsicness of my variable...what
>does "covariance does not exist" mean?...I can calculate covariance with 
ISATIS >and when variogram increases not bounding around a priori variance my
>covariance will be negative...but it continue to exist!....so how to
>distinguish second order from intrinsic variables?....and decide if beeing 
able >to use only variogram or choose between covariance or variogram?....

There is a trick in the relationship between "stationarity" and "existing of 
variogram". Here is my logic,

(1) When there is stationarity of both 1st and 2nd order, the semivariogram 
exists. On the other hand, (2) If there is semivariogrm exists, does that mean 
the stationarity of both 1st and 2nd order exist? Intrinsicness Theory is 
based on ideal situation I can not totally agree after some simulation I have 
done. The first case prevails, but not the second case.

When you have second order, or covariance, non-stationarity, it does not mean 
the covariance not exising. Rather, it means the variation is too large and 
the 2nd order stationarity does not exist. The 2nd order stationarity is also 
called homoskedasticity, while non-stationarity is heteroskedasticity (Check 
out http://www.riskglossary.com/articles/heteroskedasticity.htm)

There are two ways we can salvage 2nd order non-stationarity in my opinion:

First is about scale. If the study area is large and contains many data, 
instead of using the whole area,  conduct some cluster analysis and break the 
area into smaller scale areas, each of which may abide by intrinsic 
hypothesis. This means you will get semivariograms on each sub-region of the 
whole study area.

Second way is about transforamtion. If the data is limited or the study area 
is small, conduct some standardized transformation of data so the covariance 
will become stationary.

Hope this helps.



Shing

Shing-Tzong Lin
Teaching and Research Assistant
Department of Geography
Texas State University, San Marcos
(512)245-1935


* By using the ai-geostats mailing list you agree to follow its rules 
( see http://www.ai-geostats.org/help_ai-geostats.htm )

* To unsubscribe to ai-geostats, send the following in the subject or in the 
body (plain text format) of an email message to [EMAIL PROTECTED]

Signoff ai-geostats

Reply via email to