sorry for not writing this earlier. This is a summary of the answers to my
> can anyone tell me if I can use variogram
> > estimation for a data set of only 30, 50 or 100 measurements? And what
> > about qualitative data?
Andrew Liebhold wrote he wouldn't advise a variogram based on < 100 points,
because he doesn´t think that it is very reliable
Andrew Lister wrote, he would trust it if one had at least 20 pairs of points
in each lag
class; it is possible to weight the variogram by the number of pairs/lag class
You can use it for as little as 30 datasets - but it might be very erratic.
Tony Lolomari wrote, one will just have to use some "soft" (qualitative)
information to guide
the sill and range parameters. Depending on what algorithm you're using
(kriging, gaussian simulation, indicator simulation), you can get the
theoretical sill value (=variance, = 1, = 1(1-p) respectively.
Dragoljub Pokrajac wrote : According to my experience, it is not Ok to use
variograms for 30 data. For 100, probably you can get something useful, but
must regulate the number of variogram bins.
Mark Murphy wrote, one can certainly try to evaluate the spatial continutiy
that many measurement. Depends a lot on the data spacing and the
type of continuity one expects but he has done several exercises producing
kriged estimates (using variograms) for as few as 30 data points. It is also
possible to do categorical data by using indicator (ie yes/no coded to 1 or 0)
coding to get an idea of spatial continuity.
Carme Hervada i Sala wrote, that it is possible, but if you do them with less
than 100 data, their models are not going to be quite reliable- you cannot be
sure they represent all the population
Åsa Danielsson mentioned to check this reference:
Webster R. and M. Oliver (1992) 'Sample adequatly to estimate variograms of
soil properties'. Journal of Soil Sciences 43: 177-192. They state you need at
least 100 observations to get a reasonable variogram, but 250 would give a
good one. (that´s really an interesting paper in my opinion).
Heinz Burger wrote, it is a l w a y s possible to make a spatial analysis. If
there isn´t an autocorrelation, it´s because of missing data in this distance.
He told that a structure is to see with about 30 measurements. Qualitative
data can be coded with Indicator-Transformation (look at: Deutsch&Journel:
Edzer J. Pebesma finally wrote : Both is possible.
Thank you again to all who answered!
*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!