A detail that caught my eye:
The loop starts
forv i=1/`=_N' {
su lat in `i', meanonly
scalar plat=r(mean)
su lon in `i', meanonly
scalar plon=r(mean)
This seems needlessly indirect.
The mean of a single value is
just that value, so
forv i=1/`=_N' {
scalar plat = lat[`i']
scalar plon = lon[`i']
Nick
[email protected]
Austin Nichols
> Assuming you have a dataset of legal providers, you don't need to
> treat people differently depending on what type of state they live in,
> since the nearest provider for some people cannot lie inside their
> state, while it may for others, but the nearest provider *can* lie
> outside the state for either type of person. Assume you've got N obs,
> one per person, with lat/lon in your primary dataset people.dta, and M
> providers with lat/lon in your provider data prov.dta. Then try this
> approach on for size (assumes you may want to know something about the
> nearest provider other than distance to it, e.g. what state it's in,
> or costs of services):
>
> use people
> gen provid=.
> gen mindist=.
> sort id
> forv i=1/`=_N' {
> su lat in `i', meanonly
> scalar plat=r(mean)
> su lon in `i', meanonly
> scalar plon=r(mean)
> preserve
> use prov, clear
> vincenty lat lon plat plon, gen(dist)
> su dist, meanonly
> keep if dist==r(min)
> * What do you do if multiple providers are all closest?
> * I pick one at random.
> * You could instead: scalar n=_N and save IDs
> * in a loop, then expand n after the restore
> scalar cprovid=provid
> scalar mdist=dist
> restore
> replace provid=cprovid in `i'
> replace mindist=mdist in `i'
> }
> sort provid
> merge provid using prov, uniqusing
>
> You can get vincenty.ado via the command
> net from http://www-personal.umich.edu/~nicholsa/stata
> or use your own method.
>
> All of the above is untested code--use at own risk...
>
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
