[R] who can tell me the reason why it is different on calculating Moran's I using ARCGIS, Geoda and R?

Roger Bivand Roger.Bivand at nhh.no
Fri Mar 24 08:27:02 CET 2006

Please use the appropriate R-sig-*, here R-sig-geo:

On Fri, 24 Mar 2006, zhijie zhang wrote:

> The attachment is my dataset:
> 1.ccc.shp (the original data)
> 2.ccc.gwt, which is computed by Geoda;

Most list readers will not know (perhaps thankfully) what a shapefile is.

> Introduction to the variables in my data:
>  ID: key variable;
>  N_LATITUDE: latitude measured by GPS;
>  E_LONGITUD:  longitude measured by GPS;
>  LIVES: attribute data
> I get the different result of Moran's between ARCGIS and Geoda, R, why?
> ARCGIS:spatial statistics tools:spatial autocorrelation(Moran's I)=0.0343
> GeoDa:Moran's I=0.1539
> R:Moran I statistic =0.153905049
> The key arguments in R that i use:
> ccc<-read.gwt2nb("ccc.GWT",region.id=ID)
> ccc2<-nb2listw(ccc, glist=NULL, style="W", zero.policy=TRUE)
> moran.test(LIVES,ccc2,alternative="two.sided")
> what is the problem? I'm very confused by it?

Further, you can accept that GeoDa and moran.test() in the R contributed 
spdep package do this correctly, and the most likely reason for any 
difference is your choice of spatial weights (the same in GeoDa and R, 
different in ArcGIS). 

Please ask ESRI why their result is different! By the way, both R/spdep
and GeoDa are free to download and are (quite) easy to learn, ArcGIS is a
gigabyte install, isn't at all free, and takes months to learn. Beyond
that, R/spdep is not limited to the Windows platform.

> Thanks very very much!
> --
> Kind Regards,Zhi Jie,Zhang ,Department of EpidemiologySchool of Public
> HealthFudan UniversityTel:86-21-54237149

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

More information about the R-help mailing list