[STATSGRASS] R/grass, numerical instability/kriging

[Edzer J. Pebesma]

Well, the duplicate observation cause is the most common cause; it would 
be helpful if the software could find this out by itself. OTOH, checking 
this for huge data sets may be pretty expensive.

Once the error occurs, it's deep deep deep in the C code, and hard to 
find out from there; gstat jumps back rather crudely (using longjmp()).
--
Edzer


Tomas Lanczos wrote:
> Hello,
>
> finally I solved it - one of the points were duplicated another two points
> were too close to each others, now it's perfect, thank You. However, is
> there a common rule for e.g. selecting resolution for the given smallest
> distance the measurments?
>
> regards
>
> Tomas
>
>   
>> Tomas,
>>
>> There's a few possible of reasons; the most common ones are 
>> more than one measurement sharing the same location, or 
>> insane variogram models in kriging; anothers is perfectly 
>> correlated predictors, possibly in a local search neighbourhood..
>>
>> We need to see more from your case to find out.
>>
>> Best regards,
>> --
>> Edzer
>>
>> Tomas Lanczos wrote:
>>     
>>> Hello,
>>>
>>> how can I prevent numerical instability in kriging? Using R with 
>>> grass63, library gstat.
>>>
>>> Many thanks
>>>
>>> Tomas
>>>
>>> _______________________________________________
>>> statsgrass mailing list
>>> statsgrass at grass.itc.it
>>> http://grass.itc.it/mailman/listinfo/statsgrass
>>>   
>>>       
>
>