[R-sig-Geo] distribution of distance for lags (spdep package)

[Roger Bivand]

On Tue, 3 Oct 2006, Horacio wrote:

> Dear R-geo people,
> 
> I just discovered the beauty of the spdep package to make spatial 
> analysis of point data easier... However, as I am climbing the learning 
> curve I have a question regarding the estimation of the the lag in the 
> Moran and/or corelogram function.
> 
> As I understand, if the you generate a neighbor-list using the 
> knearneigh 
> <http://finzi.psych.upenn.edu/R/library/spdep/html/knearneigh.html> 
> function you get a 'knn' class of the points I k distance apart. Then 
> when you calculate the Moran's I (at least that's what I did) you get 
> the Lag in terms of the number of links which in practical purposes is 
> k. so, the first correlation value (od Moran's I) is the obtained by 
> comaring all the sites that have a direct link between them, then, the 
> second lag does it in terms of the ones that are 2 links apart and so on...
> 
> My question is perhaps very simple... How can I generate a list to plot 
> the distribution of distances for each lag? I thought nbdists could do 
> it, but are not sure how to go about it...
> 

You are on the right track. Here is an example:

data(columbus)
col.lags <- nblag(col.gal.nb, 4)
# to get a list of neighbour lists at increasing lags
ldists <- lapply(col.lags, function(x) nbdists(x, coords))
# to get a list of lists of distances
ldists1 <- lapply(ldists, unlist)
# to flatten the list to contain vectors of distances, as shown by:
str(ldists1)
# but note that each link is included, both i-j and j-i
sapply(ldists1, summary)
# for example will give a summary or more complicated:
dens <- lapply(ldists1, function(x) density(x, bw=0.1, from=0, to=3.1))
plot(c(0,3.1), c(0,max(sapply(dens, function(x) max(x$y)))), type="n", 
  xlab="distance", ylab="density")
lapply(1:4, function(i) lines(dens[[i]]$x, dens[[i]]$y, lty=i))

for a density plot.

Thanks for the question, I can see that summary.nb() needs a rewrite which 
would have made this a bit easier.

Roger

> 
> thank you
> 
> 
> 
> Horacio
> 
> _______________________________________________
> R-sig-Geo mailing list
> R-sig-Geo at stat.math.ethz.ch
> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
> 

-- 
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