If your actual datasets are quite large, the best way (IMO) is to do this is with a hash object.   data result; if _n_=1 then do; declare hash b(dataset:"b"); b.DefineKey("id"); b.DefineData("changeY"); b.DefineDone(); if 0 then set b; end; set a; if b.find() = 0 then y = sum(y, changey); drop changeY; run; ... View more

So if your number is not found, set it to missing. You'll need to take out the leave statement and change the output so that it only happens on a line where at least one match was made. ... View more

Follow @ChrisNZ's advice #1 and my #2. ... View more

There are several problems. You have to tell SAS which key values you're going to use to search in the hash object. Since you are searching both ID and ID2, you have to explicitly list the key values in the find() method. The length of both ID variables has to be the match the length of your ID variable in your hash object. I didn't include it below, but I set the length of ID in your market_db dataset to $8. Once you find a match, you need to tell the data step what to do. I used a output and leave statement. data want; set market_db; array phn py1-py4; if _N_=1 then do; dcl hash H(dataset:'good_numbers'); H.definekey('ID','confirm_phn'); H.definedata('confirm_phn'); H.definedone(); if 0 then set good_numbers; end; do over phn; if H.find(key:ID, key:phn) = 0 or h.find(key:id2, key:phn) = 0 then do; output; leave; end; end; call missing(confirm_phn); run; ... View more

That's just the way it was coded. From SAS's documentation: This statement suppresses a title on line n and all lines after it: (Where n is a number between 1 and 10) titlen; A TITLE statement for a given line cancels the previous TITLE statement for that line and for all lines with larger n numbers. TITLE with no number assumes TITLE1. ... View more

Hmm. That really didn't clear it up for me. I think I need to know: What the "date" is in your logic below? What you mean by "xxx"? Confirmation that you do not consider a missing end date as being less than "date".   Read row 1, if end date less than date (specific field in dataset) then xxx, read new row, if ID, SITE equal ID, SITE in row 1, read next row, else perform evaluation.   ... View more

This should be easy to accomplish with SQL. If your dataset is huge, there might be more efficient ways to do it.   data have; infile datalines; input ID :8. Index $1.; datalines; 1 A 1 B 1 B 2 A 3 A 3 B ; run; proc sql; create table want as select id from have where index='A' and id not in (SELECT id from have where Index = 'B') order by id; quit; ID 2 ... View more