[R-sig-Geo] Problem for plotting Moran's I estimates

[Roger Bivand]

On Tue, 29 Mar 2005, Chloe ARCHINARD wrote:

> Hello all, I'm a new user of R and I have used the spdep package to
> calculate Moran's I estimates. I used the dnearneigh() function and
> moran.test() function and obtained what I wanted, but now I want to plot
> my results with the p-values. I regrouped my results for the Moran's I
> like
> >m=c(m1$e,m2$e...)        #m1 contain the results of the moran.test for lag1
> >p=c(m1$p.v,m2$p.v...)
> The problem is that in mx$e there are three values (moran'I statistic,
> expectation and variance) and when I plot my results I plot all the
> three values for each lag. What should I do to choose only the statistic
> for plotting and maybe the variance for each I statistic? (For the
> p-value it's ok I have found how to differentiate the significant and
> non significant values.) I hope someone understand my question and could
> help me! Thanks a lot.

If you have the output object from moran.test()

data(oldcol)
res <- moran.test(COL.OLD$CRIME, nb2listw(COL.nb))
str(res) # shows the structure of the object
est <- res$estimate[1]
pv <- res$p.value

I suggest you make a list of test result objects for your distance lags, 
and then use sapply(mylist, function(x) c(x$estimate[1], x$p.value)) to 
extract the values (you may need t() around the result to transpose the 
output). I'll try to add an example like this to moran.test() - any 
suggestions of a well-known suitable data set with known Moran's I values 
and distance bands?

Roger

> 
> Chloé ARCHINARD
> Centre d'Ecologie Fonctionnelle et Evolutive
> (C.N.R.S.-U.M.R. 5175)
> 1919, Route de Mende
> 34293 Montpellier Cedex 5 France
> chloe.archinard at cefe.cnrs.fr
> 
> 
> 

-- 
Roger Bivand
Economic Geography Section, Department of Economics, Norwegian School of
Economics and Business Administration, Helleveien 30, N-5045 Bergen,
Norway. voice: +47 55 95 93 55; fax +47 55 95 95 43
e-mail: Roger.Bivand at nhh.no