home *** CD-ROM | disk | FTP | other *** search
- Comments: Gated by NETNEWS@AUVM.AMERICAN.EDU
- Path: sparky!uunet!uvaarpa!darwin.sura.net!paladin.american.edu!auvm!PO.CWRU.EDU!PAT
- Return-Path: <pat@slc4.INS.CWRU.Edu>
- Message-ID: <9301281529.AA24974@slc4.INS.CWRU.Edu>
- Newsgroups: bit.listserv.stat-l
- Date: Thu, 28 Jan 1993 10:29:20 -0500
- Reply-To: "Paul A. Thompson" <pat@po.CWRU.Edu>
- Sender: STATISTICAL CONSULTING <STAT-L@MCGILL1.BITNET>
- From: "Paul A. Thompson" <pat@PO.CWRU.EDU>
- Subject: Re: Interactions
- Lines: 24
-
- >Mr. Keppel is quite wrong, or, more to the point, he is engaging in classic
- >ANOVA interaction thinking. An interaction variable is the product of two
-
- Well, it seems to me that when you use an ANOVA, you engage in classic
- ANOVA interaction thinking, otherwise you would use another technique.
-
- The whole point to the ANOVA is the conditionality of the statement. If
- you have an interaction, you can't make an unconditional statement. You
- can always say something, and that is what simple main effects are for,
- to allow you to make main-effect-like statements when the interaction
- precludes making any unconditional statement. Simple main effects are
- statistically reasonable statements, usually constructed with an
- appropriate power level.
-
- If an interaction is present, there may be interesting overall main
- effect differences and there may not be. If the interaction is not
- significant, and the main effect for A is, we can always be sure that at each
- level of B, there is some interesting difference between the levels of
- A. This is the type of thinking usually done in ANOVA.
-
- --
- Paul Thompson, Ph.D. | Department of Psychiatry | (216) 844-7463
- Case Western Reserve University | Cleveland, OH 44106
- -----------------------------------------------------------------
-