Date:         Thu, 21 Aug 2008 15:17:26 +0000
Reply-To:     iw1junk@COMCAST.NET
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Ian Whitlock <iw1junk@COMCAST.NET>
Subject:      Re: Exact values from SAS variables not showing as matching

Sig,

I am too dumb to even know what the problem is.

   proc sql;
     create table testSubset as
     select i,j,(ranuni(0)*1000) as j2, sign(j-calculated j2) as s
        from test
     where i> j
     ;
   quit;

Ian Whitlock
==============

Date:         Wed, 20 Aug 2008 16:56:33 -0400
Reply-To:     Sigurd Hermansen <SigurdHermansen@WESTAT.COM>
Sender:       "SAS(r) Discussion"
From:         Sigurd Hermansen <SigurdHermansen@WESTAT.COM>
Subject:      Re: Exact values from SAS variables not showing as matching
In-Reply-To:  <62432006F5965C42BAEC4EA29286EE0507AB2D2D71@EX-
CMS01.westat.com>
Content-Type: text/plain; charset="us-ascii"

Here's an interesting variant on anomalies in matching of SAS
variables that I began writing as part of a response to another
question:

data test;
   do i=1 to 1E5;
      j=ranuni(100)*1000;
          output;
   end;
proc means data=test noprint;
   var j;
   output out=mean mean=mean;
run;
proc sql;
  create table testSubset as
  select i,(ranuni(0)*1000) as j from test
  where i> j
  ;
quit;

It took me a while to discover why the program generates apparently
incorrect results. You will likely see the reason immediately. If not,
I'll explain after others have a chance to test the program and
diagnose why it doesn't work as one might expect at first glance. S