Date:         Wed, 8 Apr 2009 13:21:28 -0400
Reply-To:     oloolo <dynamicpanel@YAHOO.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         oloolo <dynamicpanel@YAHOO.COM>
Subject:      Re: Means

On Wed, 8 Apr 2009 12:04:09 -0400, Randy <randistan69@HOTMAIL.COM> wrote:

>My data is as follows:
>
>date   time        ID         VarA
>Mar1    9:00       XA          45
>Mar1    9:00       XA          16
>Mar1    10:00      XA          34
>Mar1    10:01      XB          16
>Mar1    10:02      XD          17
>Mar1    10:05      XM          18
>Mar1    10:05      XM          19
>Mar1    10:34      XM          20
>
>I want to sum and get the average of these variables.  But every ID at the
>same time has to be grouped as one, before I take the average of these
>variables. So for example, on Mar1 XA at 9:00 has to be grouped as one, and
>so should XM at 10:05.  So the data set should first look like before I
take
>the averages.
> date   time        ID         VarA
>Mar1    9:00       XA          61
>Mar1    10:00      XA          34
>Mar1    10:01      XB          16
>Mar1    10:02      XD          17
>Mar1    10:05      XM          37
>Mar1    10:34      XM          20
>
>This operation can be performed in two data steps.  Is there a way to do it
>in one data step?
>           Randy


Just realized that OP might want avg per Time+ID group, then you can go
like this:
****************;


data _null_;
     length ID $ 2;
     if _n_=1 then do;
     declare hash h();
  h.defineKey('time', 'ID');
  h.defineData('time', 'ID', 'count', 'mean');
  h.defineDone();
  sum=0;
  do until (eof);
           set have  end=eof;
     rc=h.find();
     if rc=0 then do;
     count+1;
     mean=(count-1)*mean/count + VarA/count;
     rc=h.replace();
     end;
     else do;
        sum=VarA; count=1; mean=VarA;
     rc=h.add();
     end;
  end;
     end;
  ro=h.output(dataset: 'mean2');
  stop;
run;