creating a variable which is based on information given between first. and last. of other var

[riyaaora275]

I have a data which looks like this-

Patient code	episode	code	code2
P1	1	NS	S
P1	1	S	S
P1	1	S	S
P1	1	NS	S
P1	2	NS	NS
P1	2	NS	NS
P1	3	S	S
P2	1	S	S
P2	2	NS	S
P2	2	S	S

I have patient code , episode and code and want to create a variable like code2.

Code 2 should be S if any of the rows corresponding to an episode of a patient is S.  Like for P1 , episode 1 , since middle two rows are S all rows corresponding to P1, episode 1 should show S.

How do I write a code for this?

Would first. and last. come to use here.

 I am clueless.

[Jagadishkatam]

data have;
input Patient $ 	episode	code$;
cards;
P1 1 NS	
P1 1 S	
P1 1 S	
P1 1 NS	
P1 2 NS	
P1 2 NS	
P1 3 S	
P2 1 S	
P2 2 NS	
P2 2 S	
;

proc sort data=have;
by patient episode code;
run;

data want;
do until(last.episode);
set have;
by patient episode code;
if last.episode then code2=code;
end;
do until(last.episode);
set have;
by patient episode code;
output;
end;
run;

[Kurt_Bremser]

Assuming your dataset is sorted accordingly:

data want;
merge
  have
  have (
    rename=(code=code2)
    where=(code2 = 'S')
  )
;
by patient_code episode;
if code2 = '' then code2 = code;
run;

[RW9 · Accepted solution]

Could do something like:

proc sort data=have;
  by patient_code episode descending code;
run;
data want;
  set have;
  retain code2;
  by patient_code episode;
  if first.episode then code2=code;
run; 

It wouldn't be in your original order, but I am guessing there is other information for that (i.e. dates/times or something).  Oh, an please post test data in the form of a datastep in future, the above is not tested.