Yes, Smantha, I did: proc surveyselect data=cc out=RandOrderCC noprint method=SRS /* sample w/o replacement seed=123 */ samprate=1 /* proportion of observations in sample */ OUTRANDOM; run; That didn't give four controls for every case. I should have N cases * 4 controls. Depending on the seed, I get a different number of controls. (Also: if I have for case 1 controls 1 2 3 4 and for case 2 I also have control 2, I would like to avoid that the selection procedure uses control 2 for case 2, since I only have 4 controls for case 1. I would like to have as many controls as possible.)
... View more