Date:         Fri, 23 Apr 1999 22:21:37 +0100
Reply-To:     John Whittington <medisci@POWERNET.COM>
Sender:       "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From:         John Whittington <medisci@POWERNET.COM>
Subject:      Re: Get the Mean Value and More
Comments: To: "Paul M. Dorfman" <sashole@EARTHLINK.NET>
Content-Type: text/plain; charset="us-ascii"

At 16:13 23/04/99 -0400, Paul M. Dorfman wrote (in part):
 
>I have adjusted the DATA step to account for missing values (see John's
posting)
>and repeated your experiment under NT/233/64. After the adjustment, the
DATA has one extra conditional
>instruction:
>
>DATA _NULL_;
>   DO UNTIL(END);
>      SET A(KEEP=VAR) END=END NOBS=N;
>      IF VAR > . THEN MEAN ++ VAR;
>   END;
>   CALL SYMPUT('MEAN',LEFT(PUT(MEAN/N,BEST.)));
>RUN;
 
Paul, it's kind of you to adjust your code just for me, but I don't actually
think the 'adjustment' you have done will actually have any effect at all!
Unless I'm going mad, the adjustment you should have made to cater for
missing values should end up as something like:
 
DATA _NULL_;
   DO UNTIL(END);
      SET A(KEEP=VAR) END=END ;
      MEAN ++ VAR;
      if var > . then N + 1 ;
   END;
   CALL SYMPUT('MEAN',LEFT(PUT(MEAN/N,BEST.)));
RUN;
 
>and repeated your experiment under NT/233/64. Regardless of the width of the
>input (10, 20, 50, and 100 satellite variables), the DATA step has invariably
>come on top. After the adjustment, the DATA has one extra conditional
>instruction:
 
Both this and Peter's post were very interesting.  However, it's not
actually the situation I was debating with others (perhaps I wasn't clear
enough).  When I expressed my gut feeling that SUMMARY might possibly start
winning in some 'wide file' situations, I was talking about those situations
in which one wanted to determine the means of ALL the variables - not just one.
 
Kind Regards,
 
John
 
----------------------------------------------------------------
Dr John Whittington,        Voice:    +44 (0) 1296 730225
Mediscience Services        Fax:      +44 (0) 1296 738893
Twyford Manor, Twyford,     E-mail:   medisci@powernet.com
Buckingham  MK18 4EL, UK              mediscience@compuserve.com
----------------------------------------------------------------