I think your solution is a logical and reliable one. Your variables are in fact observations, so why not transpose them... Another approach may be putting your variables into an array and sort and deduplicate it in the array. There are papers (google sas array sort) about the quicksort algorithm implemented in arrays. Cheers, Eric ... View more

Hallo Hellind, If you are sure that the refno is present every 4th observation you can merge the dataset by itself. Here is an example (unfortunately I cannot paste in this box, so I keep it simple): v stands for value (refno in your situation). data have;    input id v; 1 . 1 . 1 . 1 4 2 . 2 . 2 . 2 8 run; data want;    merge have (drop=v) test (firstobs=4 keep=v rename=(v=_v));    retain v;    if _v NE . then v = _v;    drop _v; run; Hope this helps! Eric ... View more

I'd say (untested): proc means data=divide sum; where div3 = 1 and div5 = 1; var i; run; Cheers, Eric ... View more

Try this as a start: data divide;    do i = 1 to 1000;       div3 = (mod(i, 3) = 0);       div5 = (mod(i, 5) = 0);       output;    end; run; Hth, Eric ... View more