[Openspace] More on problems with Contiguity Weights from Dissolved Shapefiles...

[David Long]

Thank you all for your replies to my note on the openspace list about 
corrupt files and weight matrices.  Here's a brief summary of what I 
found upon further exploration.

 Kurt Paulson noted that the spatial accuracy in 3.3 is different from 
that of ArcGIS 9, thus there's potential for imperfect alignment or 
sliver polygons to arise from subsequent overlays with the ArcView 3.3 
dissolved output. 

/In response to Kurt's note I took a more careful look at my results.  
Using ArcGIS 9.1 I examined the polygons created from each of my 
dissolves and discovered that the ArcGIS  dissolve did create a tiny gap 
between the dissolved polygon and its neighbors (The gap is literally a 
minuscule fraction of a millimeter for my county data but usually 
visible at the maximum zoom level).  So, technically GeoDa is correct to 
recognize it as an island polygon.  In my data there was no shift 
apparent for the ArcView 3.3. dissolve even at the maximum zoom level.  
Looking on ESRI's help pages I found we are not the first to encounter 
this problem with ArcGIS dissolves and it was supposed to be fixed with 
the 9.1 version.
/
Al Nucci reported a similar contiguity weights problem with files 
created in Arcview, so that may not be a reliable fix.  He suggested 
using the conversion to Geodatabase and subsequent export of a new 
shapefile as way of rebuilding the topology. 

/In response to Al's note, I tried converting my corrupt ArcGIS dissolve 
file to a GeoDatabase then back to a shapefile and it appears to correct 
the contiguity problem.  I should note however that, in this instance, 
there was a shift (about 4 millimeters) that could be an issue for 
subsequent intersect or overlay functions. /

/Just to make sure my original problem was replicable, I repeated the 
dissolves in ArcView 3.3 and ArcGIS 9.1 and found the same results as I 
reported previously.  The contiguity structure for the file created in 
ArcView 3.3. looks correct, but the contiguity structure for the file 
created in ArcGIS 9.1/ /was not./

 I've hung my files online in case anyone wants to look at them or 
experiment with them further.  The files are located at:
 http://www.ssc.wisc.edu/poplab/dissolve_test.zip

Included in the zip file are the following plus associated weights files:

 *Predissolve.shp: *Original shapefile with variable /Dis_fips /which I 
used for subsequent dissolves.

*ArcGIS9_dissolve.shp:* Dissolve created in ArcGIS based on /Dis_fips/

*AV33_dissolve.shp:* Dissolve created in ArcView 3.3 based on /Dis_fips/

*Gdb_dissolve.shp:* Created by converting ArcGIS9_dissolve.shp into 
geodatabase, then exporting to shapefile using ArcGIS 9.1 with default 
settings.
 

Please drop me a note if you have other useful insights.

Dave Long
dlong at ssc.wisc.edu

 

-- 
David Long	
Applied Population Lab	
1450 Linden Drive Rm. 350	
Madison, WI 53706	
Email: dlong at ssc.wisc.edu	
Phone: 608.262.3097	


www.doctorswithoutborders.org	

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://sal.uiuc.edu/pipermail/openspace/attachments/20050908/32ea613c/attachment.html