[Openspace] Univariate Moran

[Marie-Hélène Vandersmissen]

Hello,

I used to work with  MapStat implemented by Theriault (1998) in MapInfo to 
calculate Moran's I . The approach of Theriault is similar to that Zhang 
and Griffith (1997).

Since a few months, I work with GeoDa (I find it very usefull and extremely 
well documented. Congratulations!). When I compare the Moran's I from 
MapStat  with the Moran's I from GeoDa, there's always a little difference. 
Bailey & Gatrell (1995, p. 270) mention of an adjustment of Moran's I by 
its theoretical bound. I did some tests with  a small sample and I 
performed the calculations by hand. The Moran's I performed by GeoDa 
corresponds exactly to  the one I got with the equation from Bailey & 
Gatrell while the Moran's I performed by MapStat corresponds to the one I 
got with the classical equation (without the adjustment for the range [-1.1]).
For example, the Moran's I performed by GeoDa equals -0,2979 while the 
Moran's I performed by MapStat equals -0,2557. Does it mean that the result 
of the ratio -0,2557/-0,2979 = -0,85 is similar to the usual non-spatial 
correlation coefficient?

Please, could you confirm that result and indicate if it is mentionned in 
one of your publications? (I just want to well understand the difference 
and to well explain that to my students).

With many thanks!

Marie-Helene Vandersmissen

Professeure- chercheure
Departement de geographie et
Centre d'aménagement et de développement (CRAD)
Bureau 6257, Pavillon Charles-De-Koninck
Faculté de foresterie et de geomatique
Universite Laval
Quebec
G1K 7P4
Tél.: (418) 656-2131 poste 3056
Fax.: (418) 656-3960
Site du Departement: http://www.ggr.ulaval.ca
Site du CRAD: http://www.crad.ulaval.ca