Your sample output is inconsistent in terms of keeping ID duplicates because ID=00000-000002 is duplicated while the rest are not. Assuming that it's a typo and you don't want any dupes but only the first ID encountered in the data set 2, you can proceed as follows:
data want (drop = group: rc) ;
if _n_ = 1 then do ;
dcl hash h (ordered:"a") ;
h.defineKey ("id") ;
h.defineData ("id", "grade", "sex") ;
h.defineDone () ;
do until (z) ;
set ds1 end = z ;
id = groupa_id ; grade = groupa_grade ; sex = groupa_sex ; rc = h.add() ;
id = groupb_id ; grade = groupb_grade ; sex = groupb_sex ; rc = h.add() ;
end ;
h.output (dataset:"hash") ;
dcl hash u () ;
u.defineKey ("id") ;
u.defineDone () ;
end ;
set ds2 ;
if u.check() ne 0 ;
u.add() ;
if h.find() ne 0 then call missing (grade, sex) ;
run ; If you want to keep the dupes from DS2, get rid of the hash table U and the related statements with the method calls referencing U.
