I would like to know If anyone knows a formula or a rule to automatically define the number of lags of the experimental semi-variogram.
Do you always use the same value?
Do you adapt the number of lags depending on the case at hand? (more samples -> more lags) I've been using this proportion: nLags = round(sqrt(nSamples))
In the papers/books I've researched, I couldn’t find any deterministic recommendation about this process. Its always something like "...the number of lags should not be exaggerated or else the nugget effect will be overestimated..." which doesn’t help much...