[R-sig-Geo] Spatial auto-correlation globally (sphere)

[Roger Bivand]

On Sun, 23 Mar 2014, Alex Mandel wrote:

> Has anyone seen a method for creating a neighbor list, for use in
> spatial auto-correlation tests, that treats the earth as a sphere so
> that relationships can go shortest distance. I'm trying to make a
> neighbor list of all countries in the world.

Please start by specifying the support you intend to use (centroid point 
support, polygonal support, what?). If point support, setting longlat=TRUE 
in dnearneigh() in spdep with a matrix of coordinates, or using an object 
inheriting from SpatialPoints with the appropriate CRS will give you Great 
Circle distances; polygons are more complicated. Have you looked at the 
cshapes package, also mentioned in the Spatial Task View, which appears to 
do what you want to do? It has an R Journal article:

http://journal.r-project.org/archive/2010-1/RJournal_2010-1_Weidmann+Skrede~Gleditsch.pdf

Roger

>
> Yes I realize that the relatedness may not be distance and I might
> weight on some other factor but I need the neighbor list to start.
>
> Example, US <-> Japan, China or Russia should be a line across the Pacific
>
> Perhaps I need to develop the neighbor list using some sort of social
> diagramming tool instead.
>
> Any leads would be appreciated.
>
> Thanks,
> Alex
>
> _______________________________________________
> R-sig-Geo mailing list
> R-sig-Geo at r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
>

-- 
Roger Bivand
Department of Economics, Norwegian School of Economics,
Helleveien 30, N-5045 Bergen, Norway.
voice: +47 55 95 93 55; fax +47 55 95 91 00
e-mail: Roger.Bivand at nhh.no