A new GRASS program to calculate autocorrelation statistics,
v.autocorr, is ready for testing. It is available via anonymous ftp from
pasture.ecn.purdue.edu:pub/mccauley/v.autocorr.shar.Z
It will probably be put on moon.cecer.army.mil in the very
near future.
I have tested it based upon examples in:
Daniel A. Griffith, (1987). Spatial Autocorrelation - A Primer.
Association of American Geographers.
and it seems to be working fairly well. It works only for GRASS
binary vector files labeled with numeric attributes.
****To those who know what spatial autocorrelation is:****
This program calculates a Geary Ratio and Moran Coefficient, standard
errors of each under both randomization and normalization, estimated
population means (Monte Carlo simulation) for both stats under both
assumptions, and Z statistics for testing Ho: spatial
autocorrelation exists. It also optionally outputs a connectivity
matrix.
Planned enhancements include:
o calculation of Wilk-Shapiro statistics at various places in the code
(e.g., during Monte Carlo simulation, to make sure that I am sampling
enough to get a good estimate of population means).
o calculation of population sizes (perhaps someone might need this for
something)
o calculation of some measure of sparseness (suggestions?) for the
connectivity matrix.
****To programmers:****
There's some nifty statistical functions from statlib which I have
ported to K&R C (approx 32 different flavors of random number
generators) included in this code. Deviates can be calculated from
Beta, Chi-square, Exponential, F, Gamma, Multivariate Normal, and many
other distributions. Documentation for these is included.
--
James Darrell McCauley Department of Ag Engr, Purdue Univ
mccauley@ecn.purdue.edu West Lafayette, Indiana 47907-1146
** "Do what is important first, then what is urgent." (unknown) **
P.S. due to time constraints, I regret that I cannot mail to those
without access to anonymous ftp. Try asking a friend if you
fit into this category.