gouchangjiang / opengeoda

Automatically exported from code.google.com/p/opengeoda
GNU General Public License v3.0
0 stars 0 forks source link

Regression: Bug with W creation #222

Closed GoogleCodeExporter closed 9 years ago

GoogleCodeExporter commented 9 years ago
There seems to be a bug when you create weights from within the regression 
dialog - you get altered OLS results and false spatial diagnostics. I just 
replicated the problem using queen weights (with the dataset I sent 
separately). If W is generated first and then the regression with spatial 
diagnostics is run, everything's fine. If W is generated as part of the 
regression process, you get funky results as in the attached results example.

Original issue reported on code.google.com by jkoschin...@gmail.com on 13 Nov 2012 at 6:15

GoogleCodeExporter commented 9 years ago
Can't reproduce with Columbus data set.  Will try with original reporter's data 
set now.

Original comment by mmcc...@gmail.com on 13 Nov 2012 at 9:33

GoogleCodeExporter commented 9 years ago
Can reproduce with original data set.  The bug appears to be related to 
internal Regression Dialog properties.

Original comment by mmcc...@gmail.com on 13 Nov 2012 at 9:55

GoogleCodeExporter commented 9 years ago
The problem is caused by inserting (or removing) Table columns after opening 
the Regression Dialog.  This issue came up with this data set because it was 
necessary to add a new ID variable from within the Weights Creation Dialog.  
This resulted in a new variable being added to the Table and column numbers to 
be missmatched.  A check has now been added for this.  This is fixed in version 
1.3.11

Original comment by mmcc...@gmail.com on 19 Nov 2012 at 11:52