Re: st: m:m merge using zip codes

[Austin Nichols <austinnichols@gmail.com>]

Bryan Stuart <bastuart@umich.edu>:
Better to use exact lat/lon and compute a weighted average over Census
blocks/tracts/whatever.  There is nothing special about a zip code
boundary in defining the neighborhood of a prison, right?  See also
page 55 of http://www.nber.org/papers/w13246.pdf

To compute weighted averages, you can use an unmatched merge e.g.
http://www.stata.com/statalist/archive/2007-01/msg00098.html
http://www.stata.com/statalist/archive/2009-07/msg00261.html
http://www.stata.com/statalist/archive/2009-09/msg00473.html
http://www.stata.com/statalist/archive/2009-09/msg00493.html
http://www.stata.com/statalist/archive/2011-06/msg00585.html
or do the same thing in Mata (each dataset a matrix), which is faster.

On Mon, Jun 11, 2012 at 11:45 AM, Bryan Stuart <bastuart@umich.edu> wrote:
> Hello,
>
> I have two data sets. In one, each row represents a prison. Each prison has
> a zip code, but there exist some zip codes with multiple prisons. The other
> data set (from geocorr) maps zip codes into PUMAs. Some zip codes map into
> multiple PUMAs. Ultimately, I want to connect each prison to a PUMA. Zip
> codes are not unique identifiers in either data set.
>
> An m:m merge is undesirable here because it isn't consistent. Simply
> appending the datasets together (and then filling in the missing columns)
> isn't ideal either, as some prison zip codes are not in the geocorr dataset
> (because they are located in rural areas, the Census Bureau doesn't assign
> zip codes to some areas).
>
> Any ideas on how to combine these datasets? Thanks!
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/