You can get the equivalent of nodupkey sort with SAS/SQL, or at least something close. You just need to get SQL to drop extra rows at random. For example:   proc sort data=sashelp.cars out=ncars nodupkey; by make; run; proc sql; create table ucars(drop=rnd) as select *, rand("uniform") as rnd from sashelp.cars group by make having rnd=min(rnd) order by make; but why?   ... View more

The proc sort steps will find the first observation that has the maximum value of FLAG within each group defined by PROMO. In PROC SQL the best you can do is find ALL observations that have the maximum value of FLAG within each group. proc sql noprint; create table new1 as select * from a group by promo order by promo having flag=max(flag) ; quit; In ORACLE or other external databases that have newer implementations of SQL syntax you can use ROW_NUMBER to find the first observation. So something like this. proc sql noprint ; connect to oracle ..... ; create table new1 as select * from connection to oracle (select * from a order by promo having 1=row_number() over (partition by promo order by promo, flag desc) ) ; quit; ... View more