hi ... if you create a file named AUTOEXEC.SAS in the folder where SAS.EXE is located, all statements in that file are executed at startup one advantage, you get another PC ... use the same file rather than customizing using SAS windows-based commands ... View more

If you don't want to restart your SAS, here is some magic strings sometimes may work for you: *))%*))*/; ;;;; %mend; options notes; run cancel; quit; proc unk; run; Haikuo ... View more

No. I could not be  at SGF2012. FriedEgg represented me to explain my paper at SGF2102. Thanks him very much. Ksharp ... View more

OK, final afterthought. It should be possible to combine all the processing into a single DATA step (if the array will fit in memory).  Defining a temporary array will speed things up: data want;    array t_totals {4999999} _temporary_;    set have (keep=x) end=done;    if _n_ > 1 then do i = 1 to _n_ - 1;      set have (keep=x rename=(x = lagged_x)) point=i;      T = _n_ - i;      t_totals{T} + (x - &mx) * (lagged_x - &mx) / (&cx - T);    end;    if done;    do T = 1 to 4999999;        RT = t_totals{T};        output;    end;    keep T RT; run; OK, I'm pretty sure I'm done here and can't speed it up beyond that. ... View more

thanks that's really useful, it gives me some insight into how to code in SAS, really useful answer, thanks ... View more

DOW can do it. But If you need fast it ,Hash Table will be favored. But for your data, I am curious. You said you want datetime within five seconds . There two obs within 5 seconds for  26JUN11:00:42:20   0.09 26JUN11:12:42:17 3.4 26JUN11:12:42:18 3.6 You need them all, or just only need the nearest obs ? DATA product_A;    INPUT A_datetime datetime16. A_price $;    format a_datetime datetime.;    DATALINES; 22JUN11:00:00:04    1.7 22JUN11:00:00:08 1.9 22JUN11:00:00:10 2.3 22JUN11:00:05:12 2.9 22JUN11:00:05:36 3.5 22JUN11:00:32:45 3.2 22JUN11:00:39:01 3.1 22JUN11:00:39:03 2.7 22JUN11:01:10:12 2.6 22JUN11:07:22:10 2.1 26JUN11:12:42:17 3.4 26JUN11:12:42:18 3.6 26JUN11:12:42:21 3.9 ; RUN; DATA product_B;    INPUT B_datetime datetime16. B_price $;    format b_datetime datetime.;    DATALINES; 22JUN11:00:00:09    0.02 22JUN11:00:05:34    0.07 22JUN11:00:39:05    0.04 26JUN11:12:42:20    0.09 ; RUN; data want; set product_b; do until(a_datetime gt b_datetime);   if b_datetime-a_datetime lt 5 and b_datetime-a_datetime ge 0 then output;   set product_a; end; run; Ksharp ... View more

See http://www.sascommunity.org/wiki/Generating_Holiday_Lists Reeza wrote: What about holidays? You could do something like the following to double check but someone might have a more sleek answer using a custom interval and intck. The following may also need to be adjusted by 1 depending on if you're including the boundary dates or not. data my_data;   set my_data;    date=datepart(Trade_date_time);   exp_date= '22-SEP-2011'd;   weeks_left= week(exp_date)-week(date);   time_left = exp_date-date;   time_left_adj = exp_date-date-(2*weeks_left); time_left_check=0; do i=date to exp_date;      if weekday(i) not in (1, 7) then time_left_check+1; end; run; ... View more

It is my two cents. data test ; format datetime datetime18.; input datetime datetime18. ; date=datepart(datetime); time=timepart(datetime); cards; 04jul2006:08:15 04jul2006:09:15 04jul2006:09:30 04jul2006:12:30 04jul2006:13:45 04jul2006:16:00 04jul2006:23:59 04jul2006:12:30 04jul2006:14:00 04jul2006:16:00 04jul2006:16:35 04jul2006:18:35 06jul2006:09:00 06jul2006:09:20 06jul2006:09:35 06jul2006:12:00 06jul2006:12:30 06jul2006:14:21 06jul2006:16:00 06jul2006:17:35 06jul2006:18:35 ; run; proc sort data=test; by date time;run; data want(keep=datetime); set test; by date; retain one two three; if first.date then do;one=0;two=0;three=0;end; if (time ge '9:15:00't) then do;if not one then output;one=1;end; if (time ge '12:30:00't) then do;if not two then output;two=1;end; if (time ge '16:00:00't ) then do;if not three then output;three=1;end; run; Ksharp ... View more