GEOSTATS: Summary 2 on the decomposition of the nugget effect
- Dear all,
two other replies I got to my question on the decomposition of the nugget
effect that I have forgotten to put in my first summary can be found
From Guillaume Larocque
" I am not a geostats expert, but I have a little problem with the definition
of the Nugget you mentioned. If we suppose that the Nugget is the value at 0
plus a very small distance, it is like saying that our model is not adequate
at the micro scale. To me, the definition of the Nugget that says NE=MV+EV is
valid only if we say the Nugget is the value at the smallest lag for which we
have pairs of observations. If we have confidence in our sampling scheme and
model, than I think we should say that the Nugget effect at a value close to
zero is only the error variance. If not, then it is really hard to distinguish
between MV and EV.
I think the Nugget is a ill-defined concept and I hope there will be a good
discussion about it on the list."
From Paulo Justiniano Ribeiro Jr
"About the nugget effect I'm surprised with you results.
You should see a difference between the estimates only at data locations.
When the nugget is regarded as being erros variance the interpolated
surface is smooth. When it is regarded as micro-variance you should see
exactly the same surface except for the data locations where you will find
some "spikes" corresponding to the data. When you split the nugget the spike
will be proportional to amout of nugget regarded as being the
He also sent me two files to illustrate his comments
- nugget.ps: plots with kriging results in a 1-D problem with
3 options for nugget: 0%, 100% and 50% of microscale variance
- nugget.example: the code in geoR/geoS to produce the plot
Thanks again to all,
Section of Earth Sciences
Institute of Mineralogy and Petrography
University of Lausanne
Currently detached in Italy
Get free email and a permanent address at http://www.netaddress.com/?N=1
*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!