Date:         Fri, 6 Jul 2001 09:37:14 -0700
Reply-To:     Pete Lund <pete.lund@NWCSR.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Pete Lund <pete.lund@NWCSR.COM>
Subject:      Re: Using weights with PROC TABULATE
Comments: To: Soumen Mukherjee <SMukherjee@MAPSNET.COM>
In-Reply-To:  <80FCDBD6BFF3D41198790050DA1627D02F1131@columbus.mapsnet.com>
Content-Type: text/plain; charset="iso-8859-1"

Hi Soumen-
  Here's a quote from Lauren Haworth's great book, "PROC TABULATE by
Example" - "...N statistics are always unweighted, whether or not you use a
WEIGHT statement.  The SUMWGT statistic adds up the values of the weights to
create a weighted number of observations."  You can then label SUMWGT as "N"
if you want.  You might want to try this - and get Lauren's book!

----------------------------------------------------------------------------
---
Pete Lund
Northwest Crime and Social Research
313-D Fifth Ave SE
Olympia, WA   98501
(360) 528-8970  - voice
(360) 280-4892  - cell
(360) 570-7533  - fax
pete.lund@nwcsr.com
www.nwcsr.com
----------------------------------------------------------------------------
---



-----Original Message-----
From: SAS(r) Discussion [mailto:SAS-L@LISTSERV.UGA.EDU]On Behalf Of
Soumen Mukherjee
Sent: Friday, July 06, 2001 9:09 AM
To: SAS-L@LISTSERV.UGA.EDU
Subject: Using weights with PROC TABULATE


I am trying to tabulate survey responses to a standard 10-point scale
(satscore) using the following code:

   proc tabulate;
      class gender race satscore;
      table satscore*(n pctn<satscore>), all="Total" race=" "*gender=" "
            / box=' ' ROW=FLOAT;
      weight wt;

I have a sampling weight (wt) for each respondent that allows me to match
the sample to key population demographics. I would like to use this weight
when computing the count and percentage of responses at each level of the
10-point scale. Unfortunately, PROC TABULATE doesn't appear to apply weights
to the "n" and "pctn" stats. I know I can get weighted results with analyses
variables (using the "mean", "std", etc. stats). I feel like I am missing
something obvious here. Is there a way to make TABULATE print the weighted
counts and percentages?

Soumen.