Art,   Thanks for pointing out a couple of holes in my code. Your solution is slick and provided yet another way to utilize arrays.   Just for the sake of closure, I have modified my code to address the multiple words as body parts.   data query (keep= id visit _part rename=(_part=body_part)); set ink; by id visit body_part; length cumul_body_parts parts_curvis $100; retain cumul_body_parts parts_curvis ''; if first.id then cumul_body_parts = ''; if first.visit then parts_curvis = ''; if index(cumul_body_parts, strip(body_part))=0 then cumul_body_parts = catx(', ',cumul_body_parts,body_part); if index(parts_curvis, strip(body_part))=0 then parts_curvis = catx(', ',parts_curvis,body_part); if last.visit then do i=1 to countw(cumul_body_parts,','); x = strip(scan(cumul_body_parts, i, ',')); if index(parts_curvis, strip(scan(cumul_body_parts, i, ',')))=0 then do; _part = strip(scan(cumul_body_parts, i, ',')); output; end; end; run; ... View more

I believe the following code does what you are expecting.   data ink; input Id $ Visit $ body_part $6-18; datalines; a w1 Ankle a w1 Wrist a w2 Ankle a w2 Wrist a w2 Neck a w3 Wrist a w3 Neck a w3 Shoulders a w4 Ankle a w4 Neck a w4 finger spaces ; run; proc sort data=ink; by id visit body_part; run; data query (keep= id visit _part rename=(_part=body_part)); set ink; by id visit body_part; length cumul_body_parts parts_curvis $100; retain cumul_body_parts parts_curvis ''; if index(cumul_body_parts, strip(body_part))=0 then cumul_body_parts = catx(', ',cumul_body_parts,body_part); if first.visit then parts_curvis = ''; if index(parts_curvis, strip(body_part))=0 then parts_curvis = catx(', ',parts_curvis,body_part); if last.visit then do i=1 to countw(cumul_body_parts); if index(parts_curvis, scan(cumul_body_parts, i))=0 then do; _part = scan(cumul_body_parts, i); output query; end; end; run;   ... View more

Just to confirm, you don't see the title anywhere: in the graphical image (GTITLE) or outside the graphical image (NOGTITLE)? ... View more

Hatsie, There are so many good suggestions here and the unanimous theme is that you probably should leave the data as is (vertical) or convert it into a normalized structure. If I were given this data and if I understand your requirement correctly, I would approach it this way - to get the daily activity sums. You can employ various SAS procedures to analyze it further. proc sql;   create table summ as   select distinct Participant, datepart(time) as Day format=date9., sum(steps) as Sum_Steps, sum(standups) as Sum_Standups <, and so on>   from guysmiley   where <conditions to eliminate the records when the participant is sleeping>   group by Participant, datepart(time); quit; ... View more