Adding a "keep" option to the -sample- function

[Michael Costello]

Hi all,

I regularly draw samples from populations for conducting surveys and one thing that has always confused me is Stata's lack of a option to keep all the data and simply mark the selected values as 1's (and the unsampled values as 0's). Is there a reason this was left out? If I roll my own modified version of this command, and I going to create some problem that Stata long ago sidestepped by not creating this functionality?

[Clyde Schechter]

Only the developers at Stata could tell you why they chose this design for -sample-. Suffice it to say, however, that the workaround is very simple. After creating the sample, just -merge- that back to the original sampling frame data set. The sample will then be those observations for which _merge == 3, and the _merge == 1 observations are the rest. (There should be no _merge == 2 observations.)

[Mike Lacy]

I like the user-written module -gsample- (SSC), which can do what is requested, among many other nice things.

sysuse auto
gsample 20, wor gen(insample)
tab insample

[Nick Cox]

It's been a while, but this might help too.

. search swor, sj historical

Search of official help files, FAQs, Examples, and Stata Journals

SJ-5-1  dm86_1  . . . . . . . . . . . . . . . . . . . Software update for swor
        (help swor if installed)  . . . . . . . . . . . . . . . . .  N. J. Cox
        Q1/05   SJ 5(1):139
        swor rewritten to use sort, stable for reproducibility and
        sortpreserve for compatibility; help file also modernized

STB-59  dm86  Sampling without replacement: abs. sample sizes & keeping all obs
        (help swor if installed)  . . . . . . . . . . . . . . . . .  N. J. Cox
        1/01    pp.8--9; STB Reprints Vol 10, pp.38--39
        samples without replacement a specified sample size;
        alternative to sample