Date:         Mon, 22 Jan 2007 10:48:14 -0500
Reply-To:     Venky Chakravarthy <swovcc@HOTMAIL.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Venky Chakravarthy <swovcc@HOTMAIL.COM>
Subject:      Re: means first. -- wonder if this is possible in a single
              datastep
Comments: To: Hari Nath <hari_s_nath@YAHOO.COM>

I would calculate the means using PROC MEANS or SUMMARY. It is somewhat
unusual to take the means for visit 1 only.

proc summary data = test nway ;
  class visit ;
  var var1 var2 ;
  by pat treat ;
  output out = visit1 (drop = _:)
         mean = var1 var2 ;
  where visit = 1 ;
  format var1 var2 8.2 ;
run ;

data result ;
  set test (where = (visit^=1)) visit1 ;
  by pat treat visit ;
run ;

Venky Chakravarthy

On Mon, 22 Jan 2007 10:04:05 -0500, Hari Nath <hari_s_nath@YAHOO.COM> wrote:

>Hi all,
>  iam trying to get the means for only visit 1 and then trying to keep
>this in a same datastep.....can someone point me in right way...
>Many thanks
>hari
>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>data test ;
>    input pat $ treat $ visit var1 var2 ;
>    cards ;
>    1000 A 1 100 90
>    1000 A 1  70 50
>    1000 A 1  30 50
>    1000 A 2  20 30
>    1000 A 2  50 60
>    1000 A 2  40 20
>    1000 A 3  50 30
>    1000 A 3  60 30
>    2000 B 1  80 90
>    2000 B 1   .  .
>    2000 B 1   .  .
>    2000 B 2  30 30
>    2000 B 2  40 60
>    2000 B 2  80 20
>    2000 B 3  30 30
>    2000 B 3  40 30
>    ;
>run ;
>
>proc sort data=test ; by pat treat visit ; run ;
>
>data test2 ;
>        retain var1_b var2_b ;
>    set test ;
>    by pat treat visit ;
>
>    if visit=1 then do ;
>
>
>        if first.visit then do ;
>            var1_b = 0 ;
>            var2_b = 0 ;
>            counter=0;
>        end ;
>            var1_b = sum(var1 + var1_b) ;
>            var2_b = sum(var2 + var2_b) ;
>            counter + 1 ;
>
>        if last.visit then do ;
>            var1_b = round(var1_b / counter,.01) ;
>            var2_b = round(var2_b / counter,.01) ;
>        end ;
>
>    end ;
>
>    else if visit ^= 1 then do ;
>        var1_b = var1 ;
>        var2_b = var2 ;
>        counter=0 ;
>
>  end ;
>/*  if last.visit ne . then output ;*/
>run ;
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>
>current output ;
>
>pat treat visit var1 var2 var1b var2b
>
>    1000 A 1 100 90  100     90
>    1000 A 1  70 50  170    140
>    1000 A 1  30 50   66.67  63.33
>    1000 A 2  20 30   20     30
>    1000 A 2  50 60   50     60
>    1000 A 2  40 20   40     20
>    1000 A 3  50 30   50     30
>    1000 A 3  60 30   60     30
>    2000 B 1  80 90   80     90
>    2000 B 1   .  .    .      .
>    2000 B 1   .  .    .      .
>    2000 B 2  30 30   30     30
>    2000 B 2  40 60   40     60
>    2000 B 2  80 20   80     20
>    2000 B 3  30 30   30     30
>    2000 B 3  40 30   40     30
>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>
>desired output ;
>
>pat treat visit var1 var2 var1b var2b
>
>    1000 A 1  30 50   66.67  63.33
>    1000 A 2  20 30   20     30
>    1000 A 2  50 60   50     60
>    1000 A 2  40 20   40     20
>    1000 A 3  50 30   50     30
>    1000 A 3  60 30   60     30
>    2000 B 1  80 90   80     90
>    2000 B 2  30 30   30     30
>    2000 B 2  40 60   40     60
>    2000 B 2  80 20   80     20
>    2000 B 3  30 30   30     30
>    2000 B 3  40 30   40     30
>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>