Save results of distinct command to dataset

[paulvonhippel]

I'm running the distinct command repeatedly on different subsets of my data. I'd like to create a second dataset that stacks the results of the distinct command. Here's a very simple example, which counts the makes and models of foreign and domestic cars.

sysuse auto, clear
split make, limit(2)
drop make
rename make1 make
rename make2 model

display "All cars"
distinct make
display "Foreign cars"
distinct make if foreign
display "Domestic cars"
distinct make if !foreign

Instead of having the results in the output window (or in addition), though, I'd like to create a new dataset that looks something like this:

Subset	Makes	Models
All cars	23	74
Foreign cars	12	22
Domestic cars	11	52

What's the most efficient way to do this? Thanks!

[paulvonhippel]

P.S. The real problem is more complex, with more subsets and more than two levels (make, model).

[Stephen Jenkins]

-distinct- (by Gary Longton; SSC) saves its results in r(). I'd use -postfile- and its siblings to loop around the subsets you're looking at, and accumulate the results for each subset in a new dataset. You can then read that in and do whatever you want with it.

[Nick Cox]

The source for distinct is better given as the Stata Journal: At the time of writing the latest version of the code should be downloaded from SJ 15-3, but the longest write-up was in SJ 8-4.


SJ-15-3 dm0042_2 . . . . . . . . . . . . . . . . Software update for distinct
(help distinct if installed) . . . . . . N. J. Cox and G. M. Longton
Q3/15 SJ 15(3):899
improved table format and display of large numbers of
observations

SJ-12-2 dm0042_1 . . . . . . . . . . . . . . . . Software update for distinct
(help distinct if installed) . . . . . . N. J. Cox and G. M. Longton
Q2/12 SJ 12(2):352
options added to restrict output to variables with a minimum
or maximum of distinct values

SJ-8-4 dm0042 . . . . . . . . . . . . Speaking Stata: Distinct observations
(help distinct if installed) . . . . . . N. J. Cox and G. M. Longton
Q4/08 SJ 8(4):557--568
shows how to answer questions about distinct observations

[paulvonhippel]

Thanks to Stephen Jenkins for suggesting the post commands. Here is the code I wrote for the toy job:

sysuse auto, clear
split make, limit(2)
drop make
rename make1 make
rename make2 model

postfile counts str20 subset int makes int models using counts, replace
distinct make
post counts ("All cars") (`r(ndistinct)') (`r(N)')
distinct make if foreign
post counts ("Foreign cars") (`r(ndistinct)') (`r(N)')
distinct make if !foreign
post counts ("Domestic cars") (`r(ndistinct)') (`r(N)')
postclose counts

use counts, clear
list

[Stephen Jenkins]

Apologies (especially to Nick) for not getting the provenance of -distinct- correct.

[Nick Cox]

Stephen Jenkins Thanks, and fine by me. In fact Gary did more of the coding and I did more on the paper so that is reflected in the order of the authors in the help file and in the paper.