home *** CD-ROM | disk | FTP | other *** search
- Comments: Gated by NETNEWS@AUVM.AMERICAN.EDU
- Path: sparky!uunet!paladin.american.edu!auvm!PSUVM.BITNET!DMR
- Message-ID: <STAT-L%92111601021854@VM1.MCGILL.CA>
- Newsgroups: bit.listserv.stat-l
- Date: Mon, 16 Nov 1992 00:59:00 EST
- Sender: "STATISTICAL CONSULTING" <STAT-L@MCGILL1.BITNET>
- Comments: 208 Cedar Bldg.
- Comments: University Park, PA USA
- Comments: AC 814-863-2401
- From: Dennis Roberts <DMR@PSUVM.BITNET>
- Subject: So, why assumptions?
- Lines: 18
-
- Most of the comments re: testing assumptions for various statistical tests
- have argued against FORMAL testing but rther, graphical or other less
- "formal" ways of checking assumptions. It seems to me that this means that
- some thumb rule that one person adopts to say that "it doesn't violate it
- ENOUGH" will be different than some other person's who uses either a
- different way of checking or a different thumb rule criterion. I would like
- to suggest that either an assumption is important or it isn't. Simulations
- have shown us some of these cases, some where it does matter and some where
- it doesn't. I think that if we adopt the strategy of informal checking, we
- really have not solved the problem ... it still is too easy to use a method
- or adopt some criterion whereby we rarely, if ever, violate the assumption
- enough so as to discard the resulting analysis. While I am not suggesting
- that I am a formal devotee of rigid assumption testing, I would offer the
- following: if we believe that some assumptions are important, and advocate
- checking, then we should AS A STATISTICAL COMMUNITY, decide which
- assumptions are worthy of this and agree to some COMMON THUMB RULE that we
- can all use in a common way. Without this, who decides where to draw the
- line? I don't think "common sense" is good enough.
-