Tomas et al.,
This is an issue that we addressed in our (Hong et al.) article "Spatial
Analysis of Precision Agriculture Treatments in Randomized Complete
Blocks: Guidelines for Covariance Model Selection," Agron. J.
97:1082–1096 (2005), in which we elaborated a model sequence approach
for assessing and accounting for spatial correlation within the ANOVA.
While the article deals explicitly with experimental treatment
structures in randomized complete block designs, the principles detailed
hold for the type of problem that you pose. The modeling approach
utilizes the mixed model autocorrelation analysis tools that are
available within SAS PROC MIXED.
Tomas Hlasny wrote:
Dear list,
I `d appreciate your help. I have 2000 randomly distributed point data
with spatially autocorrelated attributes. The region is divided into
several zones (seamless). I need to evaluate whether the borders
between neigboring zones are relevant, i.e. whether there is a
significant difference between central values of the attributes in
adjacent zones (ANOVA problem, but samples are not independent).
Polygonal declustering is good to solve under/over sampled locations
problem, but this does not accounts for autocorrelation.
Correlation of spatially autocorrelated data has been discussed many
times, however I cannot find references on something like ANOVA of
autocorrelated data (autocorrelation within groups).
Any help appreciated, maybe also some references to relevant articles.
Thank`s a lot
Regards
Tomas
--
Jeffrey G. White, Ph.D.
Assistant Professor
Dept. of Soil Science
3207 Williams Hall
North Carolina State University
Campus Box 7619
Raleigh, NC 27695-7619
Tel: 919-515-2389 Fax: 919-515-2167
email: [EMAIL PROTECTED]
