You can use by-group processing and first. to perform this task: proc sort data=temp; by ID Year; run; Data temp; Retain What_I_Need; set temp; by id year; if first.ID or Q_Ratio in (0,1) then What_I_Need=Q_Ratio; else What_I_Need=What_I_Need*Q_Ratio; run; The retain statements keeps the previous value of What_I_Need, unless it is reset to Q_Ratio by a new value of ID or a Q_Ratio of 0 or 1. Hope this helps. ... View more

You could write a macro to do this: First, get the number of obs into a macro var using a null data step or Proc SQL. Then you could write a data step something like this: data %do i=1 %to &N_obs; NewData&i %end;; set OldData; %do i=1 %to &N_obs; if _N_=&i then output NewData&i; %end; run; ... View more

Here's another option: data temp ; input ID $ val $ source 8.; datalines; V1 0 1 V1 10 1 V1 100 2 V1 250 2 V2 50 1 V2 80 1 V2 500 2 V7 10 1 V7 20 1 V7 30 2 V7 99 2 ; run; proc sort data=temp; by id source; run; Data temp2; retain Arr_good Arr_Bad; length arr_good $50. arr_bad $50.; set temp; by id source; if first.source and source=1 then Arr_good=val; else if source=1 then arr_good=cats(arr_good,',',val); if first.source and source=2 then Arr_bad=val; else if source=2 then Arr_bad=cats(arr_bad,',',val); if last.id then output; drop val source; run; ... View more