[R-sig-Geo] Newbie Questions
yud at mail.montclair.edu
Wed Dec 20 20:56:13 CET 2006
Not entirely sure whether this would help. But if you want to find out whether or not there is a relationship between the soil variable(s) and the yield, won't a spatial regression be doing the job? For SPDEP, the errorsarlm() and lagsarlm() functions shall do the job quite nicely, and in GeoDa, you can do the same with few clicks.
The manuals for both packages contain nice examples and easy to follow.
Hope this help a bit.
Danlin Yu, Ph.D.
Department of Earth & Environmental Studies
Montclair State University
Montclair, NJ, 07043
email: yud at mail.montclair.edu
----- Original Message -----
From: "GRETA G. GRAMIG" <gggramig at students.wisc.edu>
Date: Wednesday, December 20, 2006 11:57 am
Subject: [R-sig-Geo] Newbie Questions
> I have limited experience with spatial analysis, but I am trying
> to learn. Perhaps someone with patience for a newbie will field my
> questions? If this is not the appropriate forum for such questions
> perhaps someone could suggest a more appropriate forum.
> I have a small dataset that consists of 128 x and y coordinates
> (latitude and longitude, on a small field scale). At each
> coordinate there are measurements for various soil properties
> along with a biomass yield measurement for each x,y.
> Via spdep, using k=4 nearest neighbors for the weights, I have
> calculated Moran's I and Geary's C for the univariate
> relationships. For most of the measured variables there is
> statistically significant spatial autocorrelation.
> I have also used Rgeo to plot some quintile scatterplots for each
> variable. Just from looking at these quintile plots I can see
> that, although the soil variables and yield are spatially
> autocorrelated, there does not seem to be a relationship between
> the patterns observed in soil variables vs. the patterns observed
> in the yield.
> In other words, I would like to know if there is a relationship
> between yield and phosphorus, taking into account the spatial
> dependence. But how can I test this? I have used GeoDa to compute
> the multivariate Moran's but I don't really understand what the
> program is doing or if this is the correct approach.
> Another thing I would like to be able to do is, for each variable,
> interpolate between measurements to create 2D maps of the response
> I've searched everywhere I could think of but can't find any clear
> (and simple) examples to follow. This site:
> http://leg.ufpr.br/geoR/geoRdoc/geoRintro.html was some help but
> still kind of beyond me.
> Can someone suggest a starting point (a book, article, or
> website) that might point me in the right direction? It would be
> especially great to have some examples using R to follow.
> I would really appreciate any advice on these matters!
> R-sig-Geo mailing list
> R-sig-Geo at stat.math.ethz.ch
More information about the R-sig-Geo