[Openspace] Program Crashes When Running Models

[Julia Koschinsky]

Troy,

GeoDa does not yet support spatial models with k-nearest 
neighbor matrices (you can actually run them with smaller 
datasets without crashing but the results are not correct). 
See the previous Openspace messages re. turning your k-
nearest neighbor matrix into a symmetric matrix using R 
(http://agec221.agecon.uiuc.edu/pipermail/openspace/2004-
July/000312.html).

We have also seen problems with running spatial models with 
distance matrices that include islands. One way to test 
whether the islands are causing GeoDa to crash is to create 
a distance weights matrix with a threshold (e.g., the 
default), which includes all points and then run the model 
again.

It is also a good idea to change the Regression.OLS output 
file name in the inital window when conducting multiple 
model runs since we've otherwise seen problems with previous 
model specifications getting carried over into later ones.

Hope this helps.
Julia

---- Original message ----
>Date: Fri, 30 Jul 2004 13:56:01 -0500
>From: "Troy Blanchard" <tcb44 at ra.msstate.edu>  
>Subject: [Openspace] Program Crashes When Running Models  
>To: <openspace at agec221.agecon.uiuc.edu>
>
>I am using Geoda to analyze data on ~2000 U.S. counties.  
Because my
>counties are not uniformly contiguous (about 40 island 
counties), I have
>experimented with k-nearest neighbor and distance based 
weights.  In both
>cases, when I try to estimate either a spatial lag or 
spatial error model,
>the program crashes.  Any suggestions would be greatly 
appreciated.
>Thanks,
>Troy Blanchard
>

>Troy C. Blanchard
>Assistant Professor, Department of Sociology
>Research Fellow, Social Science Research Center
>Email: tcb44 at ra.msstate.edu <mailto:tcb44 at ra.msstate.edu>
>Sociology
>P.O. Box C (Mail Address)/200 Bowen Hall (Courier)
>Mississippi State, MS 39762
>Voice: (662) 325-7886
>Fax: (662) 325-4564
>SSRC
>P.O. Box 5287 (Mail Address)/103 Res. & Tech. Park (Courier)
>Mississippi State, MS 39762
>Voice: (662) 325-4784
>Fax: (662) 325-7966