home *** CD-ROM | disk | FTP | other *** search
- Comments: Gated by NETNEWS@AUVM.AMERICAN.EDU
- Path: sparky!uunet!paladin.american.edu!auvm!PO.CWRU.EDU!PAT
- Message-ID: <9212161759.AA23683@slc4.INS.CWRU.Edu>
- Newsgroups: bit.listserv.stat-l
- Date: Wed, 16 Dec 1992 12:59:10 -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: uses of factor analysis (fwd)
- Lines: 45
-
- >
- >Byron Davis <byron@EDU.UTAH.USI.OSIRIS> writes:
- >>
- >> However, I must emphasize again that these failings are not the failings of
- >> statistical packages or statistical techniques. They are the failings of
- >> particular individuals (e.g., students, professors, authors, reviewers and
- >> editors). No tool once made, is safe from misuse but the answer to this
- >> problem is not to destroy all tools, it is to better educate the users.
- >>
- > True, very true. A damning illustration of just how true this can be has
- >been shown to me by a colleague in the form of a rejection letter he received
- >froma journal. He had written a letter pointing out that a paper just
- >published in this particular psychological journal was more than somewhat
- >flawed, in fact, the analysis performed was so wrong that the conclusions
- >reached by the paper (which managed to coincide with the author's know 'biases'
- >about the subject matter) were in fact the opposite of what was really going
- >on. My colleague, who wished to remain nameless but is known to a few on
- >this list, actually showed, in his letter, how the analysis *should* have been
- >performed, and what the conclusions *should* have been. I can't remember the
- >exact words, but the editor more or less said that it would be embarrassing to
- >publish the letter, seeing as it exposed some rather large holes in their
- >reviewing process. (I suppose that since the objections were mathematical
- >in nature, most psychologists would have ignored them anyway...)
- >
- Very true. Another example in psychiatry is a famous paper published in
- 1982, with author initials A and O, involving factor analysis of
- SANS and SAPS data. They did not rotate their initial extraction, and
- interpreted the unrotated loadings, which is deranged. This paper is
- almost a citation classic, and has spawned numerous sentences of the following
- sort "A and O (1982)showed that pos and neg symptoms are bipolar, but
- this was not supported by the analysis reported here". Finally,
- in 1990, A published an article (A, A, and A, 1990), in which they admit
- that the original paper was nuts. Incidentally, A is an editor of a
- MAJOR journal, and known to be vindictive. So nobody had the guts to
- publish the fact that the original was incorrectly performed.
-
- After the topic of factor analysis is finished, can we talk about
- deranged uses of SEMs (structural equation models)? Boy, talk about
- some bad stuff...
-
- --
- Paul Thompson, Ph.D. | Department of Psychiatry | (216) 844-7463
- Case Western Reserve University | Cleveland, OH 44106
- ------------------------------------------------------------------------------
- When I hear "family values," I reach for my revolver. - para. Hans Jonst
-