You can make it work using   data test; set A; where round(v_rho,0.1)=0.8; run; The issue is related to rounding and the way computers internally store numeric values, specifically floating points. See this thread for more info: Solved: Re: I meet with a problem of data values in do loo... - SAS Support Communities ... View more

Try this: data want; format ProtectedBalance dollar10.2; set have; if _N_=1 then do; retain patt1; pattern1="/[\$][,.\d]{1,}/"; patt1=prxparse(pattern1); end; call prxsubstr(patt1, case_notes, position1, length1); if position1>0 then ProtectedBalance=substr(case_notes, position1, length1); if reverse(substr(ProtectedBalance,1,1))=',' then ProtectedBalance=substr(ProtectedBalance,1, length(ProtectedBalance)-1); run; ... View more

Yes he must have transposed it a second time, or something else is going on. Perhaps he meant 'variable 9' instead of 'observation 9'? ... View more

Hey Gary can you check the code you posted here again? I don't see how you could get 9 observations AFTER you transpose a dataset with 10 records, using the id statement alone. ... View more

Why do you need to change the macro variable definition from what's been given? It is given as cert.data_in but you are changing to the physical path, which won't work with the set statement without the quotes! Try it the way I suggested in my previous post. ... View more

 i set dataset (&library) without  double quotation it didnt work for me Can you show us your log? It should work without the quotes: data work.data_out; set &library; run; ... View more

I'm assuming that you want to identify the culprit variables that's causing the duplication so that you can fix them. Below code will create the dataset you wanted, but it assumes there is max 1 additional duplicate. You may have to run this multiple times, or macrotize it however you want, if your data contains more than 1 duplicate observations for each arrangement_key. data a1 a2; set a; by arrangement_key; if first.arrangement_key ne last.arrangement_key; if first.arrangement_key then output a1; else output a2; run; proc compare b=a1 c=a2 outstats=diffs noprint; id arrangement_key; run; data diffs (keep=_VAR_); set diffs; where _TYPE_='NDIF' and _BASE_ > 0; run; ... View more

@ChrisNZ's code works. And to get the dataset NAMES used in his code, you can do this: proc transpose data=A (obs=0) out=A_t (keep=_NAME_ rename=(_NAME_=oldname)); var :; run; data NAMES; set A_t; set B (keep=b2 rename=(b2=newname)); libname='WORK'; memname='A'; run; ... View more