home *** CD-ROM | disk | FTP | other *** search
- Comments: Gated by NETNEWS@AUVM.AMERICAN.EDU
- Path: sparky!uunet!paladin.american.edu!auvm!UICVM.BITNET!U50161
- Message-ID: <STAT-L%92082023145225@VM1.MCGILL.CA>
- Newsgroups: bit.listserv.stat-l
- Date: Thu, 20 Aug 1992 22:03:21 CDT
- Sender: "STATISTICAL CONSULTING" <STAT-L@MCGILL1.BITNET>
- From: Allen Herzog <U50161@UICVM.BITNET>
- Subject: Error rate
- Lines: 18
-
- A user has the following problem. Survey data was
- collected on 200 subjects. There are 300 dichotomous variables
- and the subjects are divided into 5 professional groups. The
- variables have missing values either as a result of non-response
- or because of a skip question. The person who collected the
- data would like to make paired contrasts for each dichotomous item
- among the 5 groups (10 contrasts in all).
- The new SAS program in SAS 6.07 (Proc Multtest) permits
- the user to compute pairwise contrasts among the cells in the
- 2 x 5 table using Fisher's Exact test. It also permits several
- different possible adjustments (Bonferroni, Sidak, permutation
- and Bootstrap - with the possiblity of using a stepdown adjustment
- to these as well). With this data set (or almost any survey data
- of a similar kind), what would be the proper adjustment to make for
- this huge number of multiple post hoc comparisons?
- Allen Herzog
- UIC Chicago
- School of Education
-