Vince28 is right. Hashing is fastest..but..the larger the reference table (cons3 in your case), the larger the memory requirements are to hold the hash tables. If you're deleting from sas datasets, I find it sometimes faster to create an entirely new dataset. This also eliminates the problem with the deleted records still taking up disk space. You might can speed up processing by indexing cons3: proc sql; create index home on cons3.home; create index busi on cons3.busi; create index alt on cons3.alt; create table want as select t1.* from acct_dtl t1 left outer join cons3 t2 on t1.tlnbr=t2.home left outer join cons3 t3 on t1.tlnbr=t3.busi left outer join cons3 t4 on t1.tlnbr=t4.alt where t2.home is null and t3.busi is null and t4.alt is null; quit;
... View more
