I hate to say it, but a loop is one way out.
levels hospid, local(H)
levels doadmit, local(D)
gen nprev90 = .
foreach h of local H {
foreach d of local D {
qui count if hospid == `h' & doa < `d' & doa >= (`d' - 90)
replace nprev90 = r(N) if hospid == `h' & doa == `d'
}
}
Nick
[email protected]
Alexander Tsai
> I have a dataset of hospital admissions: admissions indexed by
> "admitid", patients indexed by "ptid", hospitals indexed by
> "hospid".
>
> +-------------------------------------------+
> | doadmit admitid ptid hospid |
> |-------------------------------------------|
> 1. | 21 Jun 91 160 007336 1 |
> 2. | 23 Jun 91 197 007695 1 |
> 3. | 28 Jun 91 419 126888 1 |
> 4. | 30 Jun 91 550 046134 1 |
> 5. | 02 Jul 91 696 015142 1 |
> |-------------------------------------------|
> 6. | 03 Jul 91 771 035194 1 |
> 7. | 05 Jul 91 906 006762 1 |
> 8. | 06 Jul 91 989 114879 1 |
> 9. | 07 Jul 91 1040 010445 1 |
> 10. | 11 Jul 91 1328 035194 1 |
> |-------------------------------------------|
> 11. | 12 Jul 91 1403 016055 1 |
> 12. | 12 Jul 91 1402 001565 1 |
> 13. | 15 Jul 91 1609 007721 1 |
> 14. | 16 Jul 91 1694 015450 1 |
> 15. | 18 Jul 91 1856 005970 1 |
> |-------------------------------------------|
> | etc... |
>
> What I would like to do is this: for each
> patient-admission, calculate
> the total number of admissions at that hospital in the
> prior 90 days.
>
> On Sep 3 '03, Kristien Verheyen posted a message to the listserv
> (subject: "Generating 'moving sum' variable") that had to do with
> calculating moving sums, but my data are a little more
> complicated. My
> first thought was to expand the data with dummy dates of
> admission and
> generate moving sums by the prior 90 rows, but then I run into the
> problem that some hospitals have multiple admissions on one day.
>
*
* 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/
