Does GeoDa's Local Moran's I use weights specified in a custom GWT file?

106 views
Skip to first unread message

Arun inDC

unread,
Nov 13, 2019, 5:51:18 PM11/13/19
to Openspace List
Hello,

I generated a custom weights matrix in ArcGIS Pro. The weights matrix is based on a threshold distance that is "relaxed" if a feature doesn't have at least 5 neighbors. In addition, inverse distance weighting is used to generate the weights. From what I can tell, GeoDa does not have the ability to generate this type of weights matrix--please correct me if I'm wrong.

GeoDa gave me an error message when I attempted to load the SWM file created by ArcGIS Pro; the error message is of the same type that people report here. So I took the SWM file and converted it to a table that I then used to create a GWT file with some judicious editing in a text editor.  The GWT file loads into GeoDa and the neighbors in the weights matrix are correctly identified. Moreover, the row-standardized weights in the third column of the GWT file look right.

I am using GeoDa to calculate Local Moran's I with EB rate statistics and my question is whether the statistics calculated will make use of the weights specified in the GWT file.  I am uncertain because the GeoDa documentation online suggests that the weights in the GWT file aren't always used by GeoDa.

Thanks for your help,

Arun






Nicolas Cadieux

unread,
Aug 13, 2020, 5:39:51 AM8/13/20
to Openspace List
Hi,
Did you get an answer on this?  I also want to make my own weight file...  Any hints on how to proceed?
Thanks
Nicolas

Luc Anselin

unread,
Aug 13, 2020, 9:13:00 AM8/13/20
to openspa...@googlegroups.com
As is well documented, GeoDa does not use the actual  weights in a distance weights file in the computation of
spatial autocorrelation statistics., Only the existence of a neighbor relation is used,  i.e., only a binary relationship.

The main reason for this is that in an exploratory approach, the weights should be kept as simple as possible, in
order to avoid putting too much structure  in the analysis.  This is different in regression analysis. The PySAL spreg
module is much more flexible in that respect.

GeoDa does have the functionality to create inverse distance weights, but the weights are not used.

L.

--
You received this message because you are subscribed to the Google Groups "Openspace List" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openspace-lis...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/openspace-list/8ee9515b-d636-4d55-84b6-2c349807856c%40googlegroups.com.

Reply all
Reply to author
Forward
0 new messages