Randomly select 4 persons from a roster

[Pina]

I would like to randomly select 4 household members from a household roster. Any example in this regard will be deeply appreciated

[sherrell]

Have you looked at the CSPro-provided random functions? Specifically,

https://www.csprousers.org/help/CSPro/r ... ction.html
https://www.csprousers.org/help/CSPro/r ... ction.html
https://www.csprousers.org/help/CSPro/v ... ction.html

The last one might be closest to what you want. You would create a valueset with the names of the HH members, and invoke the function.
Sherrell

[Pina]

Thanks so much Sherrell. The code I am using is provided below
preproc
seed(1009);
valueset respondent_valueset;

do j = 1 while j <= totocc(MODULE_2000)
if S2_2I(j) = j then
respondent_valueset.add(S2_1I(j), j);
respondent_valueset.randomize(exclude(0));
endif;
enddo;
setvalueset($, respondent_valueset);

Assuming I have more than 4 household members, the randomization works fine which is okay with me. My challenge right now is how to restrict enumerators to select only the first 4 randomized household members.

[ares]

Hi Pina,

You can generate 4 random number between 1 to N and then the one selected are the one at the different four positions. But you should memorize the selection such that it should not change when they re-open the application.

To randomly select household members, we use a Kish table. You can look for some online. Based on the size of the household and a second criteria (like the last digit of the identification number of the household) you will always get the same list of selected members if their orders remain the same

[sherrell]

Yes, a Kish grid is a very popular option, tho it will take a bit more programming to accomplish. There have been prior conversations on this on the forum, here is one from a few years ago:

viewtopic.php?f=1&t=2417

>My challenge right now is how to restrict enumerators to select only the first 4 randomized household members.

If you create a valueset using the first four persons randomly selected from the HH, the enumerator can only choose those persons. And as Ares pointed out, you only want to do this once--so you may want to store the values selected and ensure those values don't change (if using one of CSPro's random functions rather than a static Kish grid). On the other hand, If you allow persons to be dropped or added from the HH during the interview or on a later revisit, then that can cause things to change. You'll have to discuss with the survey directors/implementors if that situation is allowed and if so, how it should be handled.

Sherrell

[Pina]

Thanks Sherrell and Ares for your support