The code in the early post was my first pass at identifying the interval. But just saying 5 min interval does not mean a lot once you look at complex data. What if there is only a 3 min interval before a large jump in time? Does that matter? The likely hood that every series is not exactly 5 minutes complicates things. I think the answer lies in the following code (I just dont have the time to get further with it). Hopefully someone else in the community may have some ideas / solutions. data eur_usd2;   set eur_usd;   start = 0;   datetime = dhms(date,hour(time),minute(time),second(time));   avg = mean(bid,ask);   if _N_ = 1 then do;   start = 1;   secsince = 0;   lagavg = 0;   addsec = 0;   end;   else do;   secsince = datetime - lag1(datetime);   lagavg = lag1(avg);   end;   if secsince > 300 or addsec>300 then do;   start = 1;   addsec = 0;   sec1avg = avg;   end;   if start = 0 then do;   addsec = addsec + secsince;   end;   retain addsec sec1avg;   format datetime datetime16.; run; EJ ... View more