Sounds like you don't want to MERGE that data at all.  Use a SET statement with the INDSNAME= option to indicate which dataset contributed the observation. You could then store the name into a permanent variable and use that in conjunction with PROC TRANSPOSE to generate new variable names. data step1; set ds1 ds2 ds3 indsname=dsn; by id1 id2; dsname = upcase(scan(dsn,-1,'.')); run; proc transpose data=step1 name=variable out=step2; by id1 id2 dsname notsorted; var dob name; run; proc transpose data=step2 out=want(drop=_name_) delim=_; by id1 id2 ; id variable dsname; var col1; run; ... View more

Untested, but an idea of how to extend @Tom nice clean solution to multiple words.   %let searchWords = apple, banana, pears; data want ; set have ; fname=codename; nWords = countw("&searchWords"); infile code filevar=fname end=eof; found=0; do while (not eof and not found); input; do i=1 to nWords while (not found); str = scan("&searchWords", i, ","); if find(_infile_, str ,'i') then found=1; end; end; keep codepath found; run; ... View more