Hi again,
I'm not sure that I correctly understand your question. Do you mean using
the Euclidean distances as weights rather than a binary neighborhood? The
steps are to build the neighborhood with find.neighbor(). Use max.dist to
define the size of your neighborhood. Then create the spatial neighborhood
object with spatial.neighbor(). Then replace the "weights" column in the
spatial.neighbor object with the weights you want (e.g., 1/distance^2). Note
that the neighboorhood matrix and the spatial.neighborhood object have the
same row structure and sorting so that you can do direct replacements using
the distances from the neighborhood matrix. Note that some literature
suggests that the randomization used in Moran's I (if you don't know the
underlying distribution and use the argument nperm for a randomization test)
is not appropriate for residuals. In this case you should randomize the
original data, run the original model on the permutations and use the
residuals from that to compare against your original residuals (e.g.,
Lichstein et al. 2000).
Lichstein, J. W., T. R. Simons, S. A. Shriner, and K. E. Franzreb. 2002.
Spatial autocorrelation and autoregressive models in ecology. Ecological
Monographs 72:445-463.
HTH
Volker
----- Original Message -----
From: "Clegg, Philip" <P.Clegg@leedsmet.ac.uk>
To: <s-news@lists.biostat.wustl.edu>
Sent: Tuesday, December 21, 2004 4:27
Subject: Moran's I with distance matrix?
| To view the terms under which this email is distributed, please go to
http://disclaimer.leedsmet.ac.uk/email.htm
|
|
|
| Can anybody tell me how to use S-Plus to compute the spatial
autocorrelation
| of residuals (Moran's I) using computed euclidean distances between XY
| coordinates rather than "nearest neighbours"?
|
|