Hi Stephanie and Yingqi
Thank you for your replies. Now I only need to get the time to give
a look to your reported references!
I think, despite that in this case my problem is not spatial, your
references can
be useful, and for sure, Stephanie, let me know when your code works.
Then, Yingqi, more or less I come from the hydrogeological field so I'm always
happy to read something about groundwater.
From what I have seen, it seems that bootstrap techniques are generally
used for this kind of problem. The variant of the methods relies
mainly in the optimization algorithm chosen and in the way in which
data (sampling) configuration is changed (especially in simulated annealing).
I wondering if one could try to follow a different approach, based
on the statistical
characteristic of data, able to detect
directly those samples which are redundant from the side
of their informative content. In some way, kriging,
which gives less weight to clustered samples, goes already in this direction.
Sebastiano
At 16.50 06/03/2008, Melles, Stephanie wrote:
Hi Seba,
I noticed that there were no responses to your posting from a few weeks
back and I'm cleaning out my mailbox today, so I thought I'd give my 2
cents worth.
We are actually working on this type of problem in our group. As far as
I know, no there isn't any software to perform sampling design
optimization, but we are working on implementing generic code for
spatial simulated annealing in R to optimize sampling based on
interpolation error. The method incorporates minimizing error in both
estimating the regression parameters and the spatially correlated
residual. A couple of relevant papers are:
3. Brus, D.J., Heuvelink, G.B.M.: Optimization of sample patterns for
universal kriging
of environmental variables. Geoderma. 138, 86{95 (2007)
11. Heuvelink, G.B., Brus, D.J., de Gruijter, J.J.: Optimization of
sample congurations for digital mapping of soil properties with
universal kriging. In: Lagacherie,
P., McBratney, A., Voltz, M. editors. Digital Soil Mapping: An
Introductory Perspective,
pp. 1-17. Elsevier, Oxford (2006)
24. van Groenigen, J.W., Siderius, W., Stein, A.: Constrained
optimisation of soil
sampling for minimisation of the kriging variance. Geoderma, 87 239-259
(2004)
See also de Gruijter & ter Braak (1990) Model-Free Estimation from
Spatial Samples:... Mathematicl Geology, 22 407-415
and Brus and de Gruijter (1997) Geoderma 80, 1-44.
Cheers,
Stephane
-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On
Behalf Of seba
Sent: February 22, 2008 12:49 PM
To: ai-geostats@jrc.it
Subject: AI-GEOSTATS: software: optimization of sampling for regression
problems
Dear list members
I'm trying to find a software which permits to optimize the sampling
network (number and combinations of samples) for problems of bivariate
or multiple regression.
This is a typical problem in monitoring.
We start with N monitoring stations and we try to build a regression
models between variable A and B (in the simplest case).
Than we see that the model works very well and we are tempted to reduce
the number of monitoring stations so as to reduce costs. How many and
which samples we can remove from the starting set?
Thank you in advance.
Sebastiano
+
+ To post a message to the list, send it to ai-geostats@jrc.it To
+ unsubscribe, send email to majordomo@ jrc.it with no subject and
+ "unsubscribe ai-geostats" in the message body. DO NOT SEND
Subscribe/Unsubscribe requests to the list As a general service to list
users, please remember to post a summary of any useful responses to your
questions.
+ Support to the forum can be found at http://www.ai-geostats.org/
+
+ To post a message to the list, send it to ai-geostats@jrc.it
+ To unsubscribe, send email to majordomo@ jrc.it with no subject
and "unsubscribe ai-geostats" in the message body. DO NOT SEND
Subscribe/Unsubscribe requests to the list
+ As a general service to list users, please remember to post a
summary of any useful responses to your questions.
+ Support to the forum can be found at http://www.ai-geostats.org/
At 18.41 06/03/2008, Yingqi Zhang wrote:
Seba,
I just saw your question by reading Stephane's email. I am not sure
the work we did will be useful for you since it is not about
regression. We used Kalman filter and genetic algorithm for this
purpose. However if you are interested, let me know:
Y. Zhang, G. Pinder and G. Herrera, Least cost design of groundwater
quality monitoring networks, WRR, VOL. 41, W08412,
doi:10.1029/2005WR003936, 2005.
Yingqi
Melles, Stephanie wrote:
Hi Seba,
I noticed that there were no responses to your posting from a few weeks
back and I'm cleaning out my mailbox today, so I thought I'd give my 2
cents worth.
We are actually working on this type of problem in our group. As far as
I know, no there isn't any software to perform sampling design
optimization, but we are working on implementing generic code for
spatial simulated annealing in R to optimize sampling based on
interpolation error. The method incorporates minimizing error in both
estimating the regression parameters and the spatially correlated
residual. A couple of relevant papers are:
3. Brus, D.J., Heuvelink, G.B.M.: Optimization of sample patterns for
universal kriging
of environmental variables. Geoderma. 138, 86{95 (2007)
11. Heuvelink, G.B., Brus, D.J., de Gruijter, J.J.: Optimization of
sample congurations for digital mapping of soil properties with
universal kriging. In: Lagacherie,
P., McBratney, A., Voltz, M. editors. Digital Soil Mapping: An
Introductory Perspective,
pp. 1-17. Elsevier, Oxford (2006)
24. van Groenigen, J.W., Siderius, W., Stein, A.: Constrained
optimisation of soil
sampling for minimisation of the kriging variance. Geoderma, 87 239-259
(2004)
See also de Gruijter & ter Braak (1990) Model-Free Estimation from
Spatial Samples:... Mathematicl Geology, 22 407-415
and Brus and de Gruijter (1997) Geoderma 80, 1-44.
Cheers,
Stephane
-----Original Message-----
From: <mailto:[EMAIL PROTECTED]>[EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] On
Behalf Of seba
Sent: February 22, 2008 12:49 PM
To: <mailto:ai-geostats@jrc.it>ai-geostats@jrc.it
Subject: AI-GEOSTATS: software: optimization of sampling for regression
problems
Dear list members
I'm trying to find a software which permits to optimize the sampling
network (number and combinations of samples) for problems of bivariate
or multiple regression.
This is a typical problem in monitoring.
We start with N monitoring stations and we try to build a regression
models between variable A and B (in the simplest case).
Than we see that the model works very well and we are tempted to reduce
the number of monitoring stations so as to reduce costs. How many and
which samples we can remove from the starting set?
Thank you in advance.
Sebastiano
+
+ To post a message to the list, send it to
<mailto:ai-geostats@jrc.it>ai-geostats@jrc.it To
+ unsubscribe, send email to majordomo@ jrc.it with no subject and
+ "unsubscribe ai-geostats" in the message body. DO NOT SEND
Subscribe/Unsubscribe requests to the list As a general service to list
users, please remember to post a summary of any useful responses to your
questions.
+ Support to the forum can be found at
<http://www.ai-geostats.org/>http://www.ai-geostats.org/
+
+ To post a message to the list, send it to
<mailto:ai-geostats@jrc.it>ai-geostats@jrc.it
+ To unsubscribe, send email to majordomo@ jrc.it with no subject
and "unsubscribe ai-geostats" in the message body. DO NOT SEND
Subscribe/Unsubscribe requests to the list
+ As a general service to list users, please remember to post a
summary of any useful responses to your questions.
+ Support to the forum can be found at
<http://www.ai-geostats.org/>http://www.ai-geostats.org/
--
Yingqi Zhang
Lawrence Berkeley National Laboratory (LBNL)
Earth Sciences Division
1 Cyclotron Road, Mail Stop 90-1116
Berkeley, CA 94720
Phone: 510 495 2983
Fax: 510 486 5686
<http://esd.lbl.gov/ESD_staff/yingqi_zhang>http://esd.lbl.gov/ESD_staff/yingqi_zhang
