post reedited and asked in other post where it was solved - post to delete

CodePudding user response：

This could definitely be done more elegantly. But this should work just fine.

data have;
    input VAL1 $2. ID VAL2 $1.;
    datalines;
P1 123 P
P1 123 P
P1 123 S
S2 44 C
S2 44 S
GG 44 P
P1 58 S
P1 58 S
;
run;

proc sort data=have out=sorted nodupkey;
by ID descending VAL2;
run;

data want;
    set sorted;
    by ID;
    if VAL2 in ('P' 'C') then output;
    else if last.ID then output;
run;