[R-sig-Geo] R: Re: indicator simple kriging with varying local mean on gstat

[facrimas at libero.it]

Dear Edzer, 

Thanks for the answer, but I still have problems.....

II want to apply the indicator simple kriging with varying local mean, 
considering secondary exaustive information.
So, with the command subset I split the primary attribute belong to the 
respective class of the secondary information.

for instance:

>data.Ni1<-subset(jura.pred, jura.pred$Ni<15)

>data.Ni2<-subset(jura.pred, (jura.pred$Ni<=25 & jura.pred$Ni>15))

>data.Ni3<-subset(jura.pred, jura.pred$Ni>=25)

Then  I evaluate the percentage of value smaller than a threshold value, and 
I obtain an exaussive set of soft indicator data, defined as:

                              y(u;zk)= Prob {Z(u)<=zk : secondary information 
at u}   (test: Goovaerts - pag. 307)

with this value I calculate the residual respect at Indicator data that 
account for all available informations, and bulid the residual semivariogram.

Using that one, I want apply the simple indicator kriging, with the command

>resCd.4th.krige<-krige(data.Ni1$res.4thCd.1Ni~1, location=data.Ni1, 
newdata=grd, model = res.4thCd.1Ni.model, beta=VECTOR)

I use a local mean, but I have problems

than, I use a constant mean, and the error is:

         "gstat: value not allowed for: covariance from non-transitive 
variogram not allowed "

Theory give that in steady assumption, we have:

     C(h)=C(0)-gamma(h)

Can I put that in command krige?

Help me, please.....  I ended my ideas........................

grazie

Cristiano

>----Messaggio originale----
>Da: edzer.pebesma at uni-muenster.de
>Data: 10/05/2009 22.41
>A: "facrimas at libero.it"<facrimas at libero.it>
>Cc: <mike.elliott at openreach.co.uk>, <r-sig-geo at stat.math.ethz.ch>
>Ogg: Re: [R-sig-Geo] indicator simple kriging with varying local mean on	
gstat
>
>The simplest solution might be to just (simple) krige the residuals, and
>add them to your known varying mean.
>
>Please be aware that this "model" suggests residuals have mean-dependent
>variance (p * (1-p)), and thus using a single variogram model for the
>full set of residuals is not in accordance with theory.
>
>I guess that this problem is larger than the one you mention --
>covariances are simply derived from variograms that reach a sill by C(h)
>= gamma(inf) - gamma(h).
>--
>Edzer
>
>facrimas at libero.it wrote:
>> Dear guys,
>>
>> I want to use the indicator simple kriging with varying local mean on 
gstat.
>>
>> Do you have a script example about it?
>>
>> I do not understand which is the local mean that I have to put in the 
command 
>> of simple indicator kriging.....
>>
>>  I do not also well understand whether if it is posssible to use the 
>> semivariogram model in SK, because theory tells that it need the 
covariance 
>> function.......
>>
>> Any suggest?
>>
>> Thank you.
>>
>> Cristiano
>>
>> _______________________________________________
>> R-sig-Geo mailing list
>> R-sig-Geo at stat.math.ethz.ch
>> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
>>   
>