Date:         Tue, 31 Mar 1998 19:12:09 EST
Reply-To:     Gregory J Toland <greg_toland@ABTASSOC.COM>
Sender:       "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From:         Gregory J Toland <greg_toland@ABTASSOC.COM>
Subject:      FW: PROC SORT Question

Thanks to the following people for their solutions.  All seem to be more
efficient than what I was using.
 
Abelson, Robert
Robin R. High
Tremblay, Bernard
Zack, Matthew M.
 
rsoong@janus.jnj.com
girish.s.patel@ccmail.census.gov
 
One Solution:
Greg,
 
Assuming that MONTH is a numeric variable, a PROC SQL solution should work
well (this is untested):
 
proc sql;
     create table output as
          select distinct car, state, min(month) as month
           from input
           group by car, state;
quit;
 
If your MONTH variable is not numeric, write an INFORMAT to read it into a
numeric variable, say NMONTH.
 
HTH,
 
Bob Abelson
Otsuka America Pharmaceuticals
"A nuclear power plant is infinitely safer than eating, because 300 people
choke to death on food every year." - Dixy Lee Ray
 
> -----Original Message-----
> From: Gregory J Toland [SMTP:greg_toland@ABTASSOC.COM]
> Sent: Tuesday, March 31, 1998 11:38 AM
> To: SAS-L@UGA.CC.UGA.EDU
> Subject: PROC SORT Question
>
> I am looking to see if anyone has a better way of solving this example:
>
>         My input dataset contains three variables CAR STATE MONTH
>
>         Input:
>                 Ford    Florida         March
>                 Ford    Florida         February
>                 Ford    Florida         June
>                 Chevy   Virginia        August
>                 Chevy   Virginia        May
>                 Chevy   Virginia        December
>
>         I want my output dataset to contain the first Monthly occurrence
> of a
> CAR STATE combination.
>
>         Output:
>                 Ford    Florida         February
>                 Chevy   Virginia        May
>
>         To solve this I do the following:
>
>         Proc Sort data = Input out = interim
>            by CAR STATE MONTH;
>         Run;
>
>         Proc Sort data = interim out = output nodupkey;
>            by CAR STATE;
>         Run;
>
>         Does anyone know of a better technique?  My input dataset has
> several
> thousand records with about 100 variables.  To sort the dataset twice
> takes up a
> lot of resources.  I also perform this technique several times throughout
> my
> master program.  Help!!!!!  Thank you.
>
> Gregory J Toland
> Abt Associates Inc.
> Sr. Programmer Analyst
> greg_toland@abtassoc.com
> (301) 913-0694