topic Re: proc surveyselect in SAS Programming

proc surveyselect

Kiteulf — Wed, 16 Nov 2016 13:23:42 GMT

How can I get the proc survey select to randomly pick cases based upon the criteria that the sum of the variable income should not exceed 0.7 of the (sum of income of selection/ income for total sample)

Re: proc surveyselect

Ksharp — Wed, 16 Nov 2016 14:25:26 GMT

You are looking for variables selection method.
Check SELECT= option of MODEL statement.

Re: proc surveyselect

Ksharp — Wed, 16 Nov 2016 14:29:32 GMT

Sorry. Ignore my last post. wrong post at wrong place.

Can you post an example to explain what you require ?

Re: proc surveyselect

ballardw — Wed, 16 Nov 2016 16:31:00 GMT

Please provide the version of SAS you are using. There may be options to help with this but they may be version dependent.

Precalculate what 0.7 of the income for the total sample would be: Example: Suppose the Total income in your set is 1000000, available from many procs in different ways: means, summary, univariate or freq to name a few.

Then the target you are looking for is I think 0.7 * 1000000 or 700000.

Use 700000 as the SAMPSIZE with your Income variable as the Size variable.

Re: proc surveyselect

PGStats — Wed, 16 Nov 2016 19:48:12 GMT

The Size variable in proc SurveySelect is for PPS sampling. PPS means Probability (of selection) Proportional to Size. Use a Size variable only if sampling with PPS.

Re: proc surveyselect

PGStats — Wed, 16 Nov 2016 19:54:34 GMT

Do you want a fixed sample size with at most 70% of incomes, or do you want to randomly pick sampling units until the income reaches 70% of incomes?

Re: proc surveyselect

Kiteulf — Thu, 17 Nov 2016 07:26:16 GMT

The latter!

Re: proc surveyselect

Ksharp — Thu, 17 Nov 2016 10:01:08 GMT

How about this one:

proc sql noprint;
select sum(air)*0.7 into : sum
 from sashelp.air;
quit;

data have;
 set sashelp.air;
 rand=rand('uniform');
run;
proc sort data=have;
 by rand;
run;
data want;
 set have;
 temp+air;
 if temp lt &sum;
run;