What you want is not clear to me, and given that there are a dozen people who usually respond 10x faster than I can, I'm guessing it's not clear to others. You really should post an example of what the desired end state is. That said, here's my poke at what you want using two approaches that hopefully gets you on the right track. But if I'm way off, my apologies.     data have; input STUDYNO:32. PID:BEST. Day_1826:DATE9. Day_InitialIW:DATE9. TRICHOT_1YR:BEST12. Personal_consumption_expenditure:BEST12. Month:32. Interval:$5. TimeFrame:$20. ExpenseType:$20. Value:32.; format PID BEST. Day_1826 DATE9. Day_InitialIW DATE9. TRICHOT_1YR BEST12. Personal_consumption_expenditure BEST12.; label STUDYNO="Sample size: 55,144,41,185,191, Total:616" PID="SUBJID#" Day_1826="Day1826" Day_InitialIW="IWInitial" TRICHOT_1YR="RA 273, UR 140, RNA 80, Missing 123"; datalines; 2 486 30DEC1992 05OCT1993 1 68.892 1 Month PreResYr House 3164.5029543 2 486 30DEC1992 05OCT1993 1 68.892 1 Month PreResYr Consume 820.0784735 2 486 30DEC1992 05OCT1993 1 68.892 1 Month PreResYr Durable . 2 486 30DEC1992 05OCT1993 1 68.892 1 Month PreResYr Health 332.71026714 2 486 30DEC1992 05OCT1993 1 68.892 1 Month PostResYr House 3075.6491125 2 486 30DEC1992 05OCT1993 1 68.892 1 Month PostResYr Consume 808.7741655 2 486 30DEC1992 05OCT1993 1 68.892 1 Month PostResYr Durable . 2 486 30DEC1992 05OCT1993 1 68.892 1 Month PostResYr Health 316.74626874 ; /* Method 1 */ proc sql; CREATE TABLE want1 AS SELECT *, max((TimeFrame='PostResYr')*value)/max((TimeFrame='PreResYr')*value) AS PostOverPre format percent8.2 FROM have GROUP BY studyno, pid, expensetype ORDER BY studyno, pid, expensetype, TimeFrame='PostResYr'; quit; /* Method 2 */ proc sort data = have; by studyno pid expensetype descending TimeFrame; run; data want2; retain pre_value; set have; by studyno pid expensetype descending TimeFrame; if first.expensetype then do; pre_value = value; PostOverPre = .; end; else do; PostOverPre = value / pre_value; end; if last.expensetype then output; run; ... View more

Assumption based on the data he provided, but I agree it's better to not assume the one-to-one. I think I'll be okay without getting credit. I'll cry a bit, but I'll recover. ... View more

That's a much better and more detailed explanation...   It isn't 100% true, though. In this case, since department ID and department have a one-to-one relationship, you could just have one in the group by but add a "distinct" to the select clause.    proc sql; create table salinfo as select distinct Dept_ID, Dept_Name, mean (Salary) as AverageSalary format=comma12., min(Salary) as MinSalary format=comma12., max(Salary) as MaxSalary format=comma12. from work.combine group by dept_id; quit; ... View more

You ran the code below and got the same results?   proc sql; create table salinfo as select Dept_ID, Dept_Name, mean (Salary) as AverageSalary format=comma12., min(Salary) as MinSalary format=comma12., max(Salary) as MaxSalary format=comma12. from work.combine group by Dept_ID, Dept_Name; quit; ... View more

GROUP BY is what you need to add. ... View more

Use ODS RTF, not HTML. ... View more

prxchange will do it. This works based on your demo data, but you'll need to change the "\w..." portion if the data in the brackets has other characters, spaces, etc.   data have; length char_var $25.; input char_var $; want = prxchange('s/\(\w{1,}\)//i', -1, char_var); datalines; Apple(MacIntosh) 4.99(Each) (New)Ball2.0(Soccer) (Ketchup)Chips ; run; ... View more

I don't know how the other options you have might affect it, but at a very basic level you're using the right syntax. Try eliminating all but the order to get it to work and then add them back in to see what's breaking it.   data plot_data; do y = -2 to 1 by .2; x = ranuni(4); output; end; run; axis1 order = (-2 to 0 by .2); proc gplot data = plot_data; plot y * x / vaxis = axis1; run; ... View more