5 replies to this topic

[Ethan Welsh]

I am not a cartographer, but I'm not sure where else to turn, or even what words to use to search for more information. I have a list of zip codes with latitude and longitude information, and I'm trying to organize these into (mostly at least) non-overlapping 40 or 50 mile regions, with the end result being a list of the zip codes at the center of those regions. I can't imagine I'm the first person to travel down this road, so any help or guidance would be appreciated. We're basically trying to segment our membership database into regions. I already have an algorithm to find all of the members within x miles of a particular zip, but I'm unclear how to extrapolate this across the country without doing a search against every zip and then de-duping.

[P Riggs]

Hi Ethan. I'll try to help.

I have a list of zip codes with latitude and longitude information,

I'm not sure I understand what you are trying to do. So is this the centroid of each zip code polygon? We work with zip code health data. Zip codes are really weird. They can take any shape and can be non-continuous. If it across the country, I'm not sure what your getting that organizing by state wouldn't. Is this a smaller geographic region than country?

and I'm trying to organize these into (mostly at least) non-overlapping 40 or 50 mile regions, with the end result being a list of the zip codes at the center of those regions.

You want to partition these into square grid polygons, or is any shape OK? Is there any criteria as to how many zip codes or members should be in each partition? If square grids and no criteria, you can create a regular tessellation of square polygons, then intersect the points with the polygons. What software are you using? If ArcGIS, after creating the grid and use a spatial join to assign the zip codes to your grid. You could then create a look-up-table in your database with zip code and region number.

Maybe I am not understanding your problem? I find it easier to work with what is the end point question and work back than start with what data you have and work forward.

[loximuthal]

I don't have much to say to your problem of grouping zip codes into regions, but I can (hopefully) shed a little light on the nature of zip codes.

A zip code is not a polygon. It is a mail delivery "route", and is generally linear (though, as Philip pointed out, they may not be contiguous) but can also be points (large institutions often have their own zip code). The Census Bureau (and others, I'm sure) have attempted to come up with zip-code related polygons to do similar analysis as you are trying. Ours are called ZCTAs (Zip Code Tabulation Areas), and you can find more on them at the U.S. Census Bureau ZIP Code Tabulation Area FAQs.

An interesting national map of zip codes rendered as points is at Ben Fry's zipdecode page.

Good luck!

[pfyfield]

If I'm understanding your question, you have a list of lat-long coordinates, which you could use to create a point file. Polygons that represent an area closer to one point than to any other point are called Thiessen polygons. You can't control the size of these polygons, that's determined by the proximity of the points.

Do you have access to GIS software? This is all doable with ESRI software, but I believe Thiessen polygons require a coverage input, so you would need the full ArcInfo license (I think).

Try googling "thiessen polygon" and see if that's the direction you want to go in. Then find someone with ArcMap.

I hope that helps...

[sitesatlas]

Manifold can also do Thiessen polygons quickly and easily. They are referred to as "Voronoi areas" in their terminology; here is the program's help file entry on that: http://www.manifold....i_operators.htm

[Dennis McClendon]

Perhaps I misunderstand what you need to do, but could you just look at a map of three-digit ZIP Code prefix areas (what were once called Sectional Centers) and group them into appropriate regions?