Are you sure that the extra column wasn't one labeled '_name_' ? I dropped that one with the code I had suggested, as renamed col1 to be combination (like you did in your example). The only things you had to change in the code I suggested were the dataset names, old rather than have, and new rather than want. ... View more

The REPlicate option generates independent samples so may have the same record in more than one replicate. Some options would be to add another stratification variable if the 4 groups should have some desired feature in common such as industry or geographical region. OR cluster Or create larger sample and then assign groups later in a data step with something like: group = mod(_n_,4); ... View more

If I understood what you mean. data have; input id $ rank group $; cards; A     1     FOOD A     2     PET A     3     PET A     4     HOME A     5     FOOD A     6     FOOD A     7     FOOD A     8     FOOD B     1     PET B     2     FOOD B     3     PET B     4     FOOD B     5     FOOD C     1     FOOD C     2     FOOD C     3     FOOD C     4     FOOD C     5     FOOD C     6     FOOD C     7     PET C     8     HOME ; data want; set have; by id; if first.id then call missing(pet,home,n); if group='PET' then pet+1;   else if group='HOME' then home+1; if group='FOOD' or  pet=1 or  home=1 then do;n+1;if n lt 7 then output;end; drop n pet home; run; Xia Keshan ... View more

I don't see one observation that has flag_2=1 and flag__1=0  within the obs marked in red . ... View more