st: Distance from home to hospital [originally no subject]

[David Kantor <kantor.d@att.net>]

In response to Kelly Richardson's question about the distance between 
home and hospital:

The structure of this situation suggests a -cross- operation on the 
two datasets (persons and hospitals) -- at least in theory.
This would yield a _very_ long dataset.  But this is impractical
You might want to loop through one person at a time, joining hospital 
data; then select the nearest (or, say the nearest 5 hospitals); then 
somehow output just these (maybe using -post-).

There may be a way to do this in mata; I am not versed in mata (yet).

In plain Stata, I might do the following (a mixture of real code and 
psuedocode):

des using personfile
local N = r(N)  // num observations

forvalues j = 1 / `N' {
 use personfile in `j'
 cross using hospitalfile
  --compute the distance; call it dist --
 sort dist
 keep in 1/5  // or however may closest hospitals you want
 -- somehow save this info --
}


For "compute the distance", I leave that to you, as  my present focus 
is the data structure. Austin Nichols had some advice on that matter.

For "somehow save this info", you can...
 -post- it,
  or save it as a tempfile (one of many tempfiles)-- and later 
gather the pieces.

For the latter approach, you might do (within the loop):
 keep [some appropriate set of variables]
 tempfile t`j'
 save `t`j''

Then in a separate loop:
forvalues j = 1 / `N' {
 if `j' ==1 {
  use `t`j''
}
else {
  append using `t`j''
}

----

I hope this helps; good luck.
--David


At 09:35 PM 1/3/2007, Austin Nichols wrote:

> Kelly--
> This comes up just often enough to warrant an FAQ, but I don't know if
> there is one.  There is a program that _almost_ does what you want
> called -nearest- written by Nick Cox on 10 January 2003 and available
> by typing -ssc install nearest- but it does not calculate distance on
> a spherical or ellipsoidal model of the Earth.  I assume when you say
> x,y you mean you have signed decimal degrees lat and lon.  In that
> case, you can calculate distance using the -vincenty- program
> vincenty lat1 lon1 lat2 lon2, g(d1) h(d2)
> to generate distance variables.
>
> Don't have the -vincenty- program?  Here it is:
> net from http://www-personal.umich.edu/~nicholsa/stata
> net install vincenty
>
> If you want to calculate distance to each hospital (or each VA and
> non-VA hosp), you can merge the hospital data onto the patient data,
> and do the calculations inside a loop that executes once for each
> patient.  If you're more specific about the structure of your data,
> you will likely get more specific advice...
>
>
> On 1/3/07, Richardson, Kelly K. <Kelly.Richardson@va.gov> wrote:
> > Dear Statalisters,
> > I would like to calculate the distance between a patient's home and the
> > nearest hospital to their home. I have the x,y coordinates for the
> > patient's home and the x,y coordinates for every hospital in the
> > country. How do I get the distance to the nearest hospital? I looked at
> > the globdist command but I can't see how to enter coordinate variables
> > for both locations. For example, say I have 1000 patients across the
> > country and 5000 hospitals. How do I get one distance for each patient
> > that represents the distance to the nearest hospital? I'm not clear on
> > how my data should be constructed or which command to use. Right now the
> > data are in separate files, one for patients (n=1000) and one for
> > hospitals (n=5000).  [...]
*
*   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/