Date:         Thu, 2 May 2002 11:07:47 -0400
Reply-To:     "Fehd, Ronald J." <rjf2@CDC.GOV>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         "Fehd, Ronald J." <rjf2@CDC.GOV>
Subject:      Re: PROC COMPARE Options Questions
Comments: cc: Myra Oltsik <myra@DATASQUARE.COM>
Content-Type: text/plain; charset="iso-8859-1"

> From: Myra Oltsik [mailto:myra@DATASQUARE.COM]
> I have a macro which has two proc compares in it. One compares the
> output of proc means between 2 versions, and the other compares the
> output of proc freq between 2 versions. My problem is I'm getting too
> much information in the list, and not enough information in the output
> file. I need help with the options in order to get the following:

Examine the options out*

proc COMPARE
 base    = &BASE.
 compare = &COMPARE.
 out     = COMPARWS
 outall     /* == outBASE outCOMP outDIF outPERCENT       */
 outnoequal /* write only obs with one or more differences*/
 noprint;
 by        &IDLIST.;
 format    _all_;

Ron Fehd  the macro maven  CDC Atlanta GA USA RJF2@cdc.gov

>    1) ID variables which are in one file but not the other.
>    2) A list of ONLY the values that are different by the fuzz factor.
>
> In the second compare below, I run a data step after the proc to get
> only those records I need to look at. But since I also want a list of
> ID variables which are in one but not the other, I need the list. The
> list gives me EVERY comparison, even if the DIFF amount isn't printed.
> (In this compare, it is possible to have more than 32767 observations
> to compare, but only a few with a difference bigger than the fuzz
> factor. This prints out much more than I need.) What I'd like to have
> is the ID variables in the list, but the comparison differences in the
> output data set. Is that possible?
>
> Myra
>
> Here's the "meat" of the macro part:
>
>       proc compare listobs maxprint=32767 fuzz=.01
>          base=in3.&dsn3.readin
>          comp=in4.&dsn4.readin
>          out =&dsn4._in_means
>          ;
>          id var1;
>          var percpop means mins maxs sums;
>       run;
>
>       proc compare listobs maxprint=32767 fuzz=1
>          base=in3.sumfill_&dsn3
>          comp=in4.sumfill_&dsn4
>          out =&dsn3._in_values
>          ;
>          id var1 value;
>          var percent;
>       run;
>
>       data &dsn3._in_values;
>          set &dsn3._in_values;
>          if abs(percent) lt 1 then delete;
>       run;
>