[UW-GIS-L] Geocoding - dealing w/ tight clusters & confidentiality

[Craig Hanson]

Walker-

I worked on a project a few years ago with a very similar problem,
geocoding the addresses of children who had been tested for blood lead
levels. Instead of showing individual "houses" where tested blood lead
levels were in exceedance, we aggregated to the census unit (in our
case, the census tract). Then we discussed the issue with our
stakeholders and consulted existing literature on the level of
confidentiality relative to the number of cases and the underlying
population universe (e.g. the number of tested children), what
information was publicly available, and probabilities of being able to
derive individual households from aggregate data to define a cut-off
number or percentage below which we would not display the information.
You can see the results at http://www.leadsafehomes.info/.

Your cut-off number will depend on the specifics of your project and the
results of your research, but the initial aggregation should be pretty
simple. When you're geocoding your addresses, make sure that you set the
point offset distance (in Advanced Geometry Options?) to > 0 to get the
point on either side of your street. Then you do a spatial join with
your block/areal units to get the block/areal unit identifier for each
point, and summarize to get the counts. This presumes that your polygon
boundaries match your street geometries, at least within tolerances, so
that your points are placed correctly relative to your polygons.

Good luck!
Craig


___________________
Craig Hanson
GIS Analyst
Windward Environmental
200 West Mercer St., Suite 401
Seattle, WA 98119
206.812.5437
Fax: 206.217.0089
craigh at windwardenv.com

 

> -----Original Message-----
> From: uw-gis-l-bounces at mailman1.u.washington.edu 
> [mailto:uw-gis-l-bounces at mailman1.u.washington.edu] On Behalf 
> Of Walker Willingham
> Sent: Monday, July 10, 2006 12:46 PM
> To: UW - GIS Discussion & Support
> Subject: [UW-GIS-L] Geocoding - dealing w/ tight clusters & 
> confidentiality
> 
> I have a new geocoding project in which numbers are more 
> important than exact locations.  In fact it might be better 
> if locations were NOT TOO reliable as a means of protecting 
> confidentiality.
> 
> I will describe the ideal result I'm looking for & if any of 
> you have ideas that might get me close, or know someone who 
> might be able to make a suggestion, I'll be thankful.
> 
> I'm making maps in residential areas, and want to visually 
> indicate numbers of children in each block meeting certain 
> criteria.  The number per block is important, even if there 
> are multiples per address, but the location within the block 
> - or even the side of the street is not important.
> 
> Ideally I would like a dot per child in a regular cluster 
> centered about the mid-block, or just an integer indicating 
> that number for the block.  (I believe the scale of my map 
> will be large enough to accommodate my dots.)
> 
> Something like:
> 
> ____* *___/_________/__*******__/____***___/_____*_____/
> or
> 
> ____2____/_________/____7____/_____3___/_____1_____/
> 
> or alternately the dots could go to both sides of the 
> streets, either based on the side that people actually live 
> on, or simply to be able to cluster more dots near the middle.
> 
> I don't mind if there is a multi-step process to solve this, 
> perhaps creating a new feature class after geocoding which 
> calculates results to account for the problem that geocoding 
> multiple points too close together stacks points as 
> indistinguishable.  If any of the steps require either an 
> ArcEditor or full ArcInfo license, I would like to know that, 
> since ArcView 9.1 is what I have handy.  Perhaps there is a 
> script someone can suggest that would help.
> 
> The project is large enough with a fairly large data set that 
> I do want to avoid a process which requires a per geocoded 
> location tweaking in an application downstream from ArcGIS - 
> that is I really want to automate some symbology that will 
> work for this.
> 
> Thanks in advance,
> 
> Walker 
> 
> _______________________________________________
> Uw-gis-l mailing list
> Uw-gis-l at u.washington.edu
> http://mailman1.u.washington.edu/mailman/listinfo/uw-gis-l
>