GROUP BY

[Thalitacosta]

I have the following code, which is returning the accumulated sum of all observations. But I want the accumulated sum by date.

In "proc sql" would do:

group by date

But in this code that I have, what do I do?

data resul1_perf;

set resul_perf;

retain bad_acc good_acc;

if _n_ = 1 then bad_acc = bad;

else bad_acc = bad_acc + bad;

if _n_ = 1 then good_acc = good;

else good_acc = good_acc + good;

dif_acc = abs(bad_acc - good_acc);

FORMAT bad_acc COMMAX20.3;

FORMAT good_acc COMMAX20.3;

FORMAT dif_acc COMMAX20.3;

run;

[PaigeMiller · Accepted solution]

data resul1_perf;
    set resul_perf;
    by date;
    if first.date then do;
        bad_acc=0;
        good_acc=0;
    end;
    bad_acc + bad;
    good_acc + good;
    dif_acc = abs(bad_acc - good_acc);
    FORMAT bad_acc good_acc dif_acc COMMAX20.3;
run;

[Kurt_Bremser]

Use BY in the data step and the first. and last. automatic variables:

proc sort data=resul_perf;
by date;
run;

data resul1_perf;
set resul_perf;
by date;
retain bad_acc good_acc;
if first.date
then do;
  bad_acc = bad;
  good_acc = good;
end;
else do;
  bad_acc = bad_acc + bad;
  good_acc = good_acc + good;
end;
if last.date;
dif_acc = abs(bad_acc - good_acc);
format
  bad_acc
  good_acc
  dif_acc commax20.3
;
run;