Date:         Wed, 30 Jun 1999 15:06:58 -0400
Reply-To:     Tim Vorce <timv@TRILLIUMTEAM.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         timv@TRILLIUMTEAM.COM
Subject:      Keeping last unique values
Content-Type: text/plain; charset="iso-8859-1"

My data is sorta like the example.  What I want to do is to keep the last
unique values where a and b are the control variables, and c is the data.
So I want to keep the last unique pair of a and b.  A sort nodupkey would be
nice, but I understand that it doesn't always keep the last observation.

My complication is that I have another control variable, and I need to keep
more than one variable (c here).  So this code is looking ugly.  Anybody
know of a better way to keep the last unique values?


 data one;
  a=1; b=1; c=1;output;
  a=1; b=1; c=2;output;
  a=1; b=1; c=3;output;
  a=2; b=1; c=4;output;
  a=3; b=1; c=5;output;
  a=3; b=1; c=6;output;
  a=4; b=3; c=7;output;
  a=4; b=3; c=8;output;
run;

data two;
  set one end=last;
     la=lag1(a);
     lb=lag1(b);
     lc=lag1(c);
  if _n_ ^=1 then do;
    if a ^= la or b ^= lb then do;
      output;
    end; /* a^=lasta */
    else do;
     put a= b= c= la= lb=;
    end;
  end; /* _n_ ^= 1 */
if last then do;
  la=a;
  lb=b;
  lc=c;
  output;
 end;
run;
*proc print data=two;run;
data _null_;
  set two;
  put la lb lc;
run;



Tim Vorce                                        (248) 586-1999
tvorce@trilliumteam.com
Trillium Teamologies
SAS.  It's not just an attitude