I don't understand where you want to output the multiple datasets, but what's for sure is that you have way more steps than needed. Try something like this, which will be more efficient and easier to follow:   proc freq data=wtnodup.&first.; tables JOBBOARDID*DATE / list out=&first. (drop = PERCENT rename=(COUNT=COUNTNEW) where =(COUNTNEW gt 49)); run; data &second. (keep = COUNTNEW OLDCOUNTNEW DATE1 RATE FROM TILL JOBBOARDID RATE); set &first.; by JOBBOARDID DATE; format DATE1 FROM TILL monyy7.; DATE1 = input(cats(DATE),yymmdd8.); FROM = lag12(DATE1); OLDCOUNTNEW = lag12(COUNTNEW); if lag12(JOBBOARDID) eq JOBBOARDID and intck('month', FROM, DATE1) eq 12 then do; TILL = DATE1; RATE = ((COUNTNEW/OLDCOUNTNEW)-1)*100; output; end; run; proc means data=&second. noprint nway; class DATE1; output out=RESULT.&final. median(RATE)=MEDIANRATE; run;   Many more optimisation tips in http://www.amazon.com/High-Performance-SAS-Coding-Christian-Graffeuille/dp/1512397490 ... View more

Hi Tom, Sorry for the confusion. Actually, I would like to take the median of the the company if the company exists in both data set. So what I think about doing is first of all do a proc means to do a count for the companies because they are in character format: PROC MEANS DATA=count noprint;         class advertiser;         var date;   output out = countdata N=Count This gives me a count for each company which then I can take and divide by 2 to get the median. So if McDonald is mentioned 5 times, the count will give me 5 for Mcdonald  and then I can divide it by 2 to get a median. And I can do this for all of the companies. What's tricky for me is to calculate the growth rate of the median between only those companies that are in the dataset in Jan 05 and 06, then Jan06 and 07...in a way that if there is a company in Jan06 that and one in Jan05 then I calculate the median growth rate, but if the company that is in Jan06 and not Jan07 to ignore that company. So for example if Mcdonald is both in dataset Jan06 and Jan05, then calculate a median growth. But if McDonald is not present in Jan07 but is in Jan06,then ignore it. I can't figure out how to do that for all different months and years across different States. SO basically calculate a 12 month median growth rate for only those companies that are present in the 12 month apart months across different states. Sorry for writing too much but I thought I should explain it and clarify it as much as I can. ... View more

You can subset within PROC MEANS, using WHERE on the output data set: proc means data=count nway; class company; var date; output out=want (keep=company n where=(n > 2)) n=n; run; When you add a CLASS statement, you will need to know how NWAY works as well. Good luck. ... View more