home *** CD-ROM | disk | FTP | other *** search
- Comments: Gated by NETNEWS@AUVM.AMERICAN.EDU
- Path: sparky!uunet!zaphod.mps.ohio-state.edu!darwin.sura.net!paladin.american.edu!auvm!SUVM.BITNET!LAWORK
- Message-ID: <STAT-L%92083106482391@VM1.MCGILL.CA>
- Newsgroups: bit.listserv.stat-l
- Date: Mon, 31 Aug 1992 06:44:12 EDT
- Sender: "STATISTICAL CONSULTING" <STAT-L@MCGILL1.BITNET>
- From: Larry of Layne <LAWORK@SUVM.BITNET>
- Subject: using weighted data in SAS
- Lines: 33
-
- I am currently analyzing some financial data that I obtained
- from the federal government for my dissertation. The tape
- includes the usual variables one might expect (age of head of
- household, number of children, family income and assets) plus a
- weight variable. When the data are used with the weight
- variable, the actual sample of 767 is blown up to 8,448,705.
- ** The documentation and codebook that came with the data state
- that the weight variable must be used. **
-
- The problem with SAS is that when statistical tests such as
- chi square and gamma are run on the weighted data, they always
- turn up as "significant." I assume that this is because the
- weighted sample is so huge and SAS reads the weighted sample size
- as if it were the actual one.
-
- I have tried to deal with this problem by creating a new
- weight variable. For work with this sample, I did this by
- dividing B3016 (the original weight variable) by different
- numbers until I found one (11015.26) that produced a weighted
- total in the frequency tables that is identical to the actual one
- of 767. This "solution" involved guesswork and interpolation on
- my part and since using weighted data is not uncommon, there
- must be a more precise way of addressing the problem in SAS!
-
- Can anyone suggest how I might go about correcting for the size
- of the weighted sample so that all of the statistics I obtain
- from SAS will be accurate?
-
- Thanks for any help!
-
- --Larry Layne, Computing Services, Syracuse University --
- --120 Hinds Hall, Syracuse, NY 13244 USA (315)443-2196 --
- --Bitnet: LAWORK@SUVM Internet: LAWORK@suvm.syr.edu --
-