Tom Check the input vector.  Question appears in the input header, with null values in the data supplied. Richard in Oz ... View more

I found the logic for this very difficult to follow and execute until I understood that the effect of flag = 10 differed whether it was the first record in the group (when in effect it behaves like "other") or later in the sequence.  So I defined xflag = flag except when it was the first record and flag=10, in which case xflag = 0.  I defined snogroup to be the value of sno for the group (needed unless you do the logic in a separate data step with BY group processing), discard to be the effect on subsequent records in the group, and firstflag to be the first instance of xflag = 1 or 10 in the group. While this answer delivers the required results for the data specified it wold not vastly surprise me if there was some combination of input records where the result was not as expected. Richard in Oz Data want ; input sno A B flag; Retain snogroup 0        discard 0        firstflag 0        ; xflag = flag ;       If NOT(snogroup = sno)   Then     Do ;       snogroup = sno ;       discard = 0 ;       firstflag = 0 ;       If flag = 10         then xflag = 0 ;     End ;   Select (xflag) ;     When (1)       Do ;       If Firstflag = 0         Then           Do ;             Firstflag = 1 ;             Output ;             Discard = 1 ;           End ;       If Firstflag = 10         Then           Do ;             Discard = 1 ;           End ;       End ;      When (10)       Do ;       If Firstflag = 0         Then           Do ;             Firstflag = 10 ;           End ;       If Firstflag = 10         Then           Do ;             Discard = 1 ;           End ;       End ;            Otherwise         Do ;           If Discard = 0             Then Output ;         End ;   End ;   Drop snogroup Discard Firstflag xflag ; cards; 1 0 1 . 1 1 1 1 1 10 2 10 1 1 1 1 1 10 2 10 2 0 1 . 2 10 1 10 2 1 1 1 2 10 1 10 2 1 1 1 3 10 1 10 3 99 1 . 3 1 1 1 3 10 2 10 4 1 1 1 4 0 1 . 4 10 2 10 ; sno A B flag 1 0 1 . 1 1 1 1 2 0 1 . 3 10 1 10 3 99 1 . 3 1 1 1 4 1 1 1 ... View more

Presumably you know in advance which are the columns you want to use, so I would suggest refining PGstats response %Let QVars = Q1 Q2 Q3 Q4 Q5 Q6 Q7 Q10 Q7B Q11 Q9 /* ... etc */ ; data want (drop = &Qvars) set MyData; array myQ{*} &QVars ; do i = 1 to dim(myQ);      question=vname(myQ{i});      score = myQ{i};      output;      end; run; Richard in Oz ... View more

Sorry - left out the by statement  - not able to check! Assuming all your data is sorted by StdPseudoId and this is a unique key to the data then you need %Macro NextYear(yr) ;   Data Want ;      Update want                 Demog&yr ;      By StdPseudoId ;   Run ; %Mend ; ... View more

PG array myQ{*} Q:; is going to include QUESTION and any other variable beginning with Q and will fail if any of them are character Richard in Oz ... View more

What about if there are less than 3 representatives of a group? Would you want the lowest salary out of 2 or the only salary represented in your dataset? If so, no problem, modify Ksharp's code slightly data want; set x; by empcode; if first.empcode then n=0; n+1; if n=3      or n<3 and last.empcode         then output; run; ... View more