Use proc summary with class group grade; Then remerge the results with your dataset, and set variables according to def. If you want fitting code, please post your dataset in a data step so we can fit the code to the data. ... View more

@jopo12345678 wrote: It works but it doesnt change to numeric format as need it.     The "calendar" icon beside the variable name clearly tells us that this is a numeric variable. If you want to get the integer number 201401 (which is basically useless for any type of computation), you need to do recorddate = input(put(recorddate,yymmn6.),6.); format recorddate 6.; ... View more

Here's an alternate solution that doesn't involve the dates.  For it to work, the dates have to be complete (no months skipped in the middle) and the observations have to be in order.  It just flags "R" observations and up to 12 observations following each "R" observation.   data want; set have; by id; if first.id then r_count=20; if rat='R' then r_count=0; r_count + 1; if r_count <= 13 then def_flag=1; else def_flag=0; drop r_count; run;   It wasn't 100% clear whether you wanted to flag the "R" record plus 12 additional records (which this does) or only 11 additional records (which your example illustrates).  But it's easy enough to change the "13" to "12" in this program if need be.   ... View more