home *** CD-ROM | disk | FTP | other *** search
- Comments: Gated by NETNEWS@AUVM.AMERICAN.EDU
- Path: sparky!uunet!zaphod.mps.ohio-state.edu!swrinde!gatech!paladin.american.edu!auvm!BEN.DCIEM.DND.CA!MMT
- Return-Path: <@VMD.CSO.UIUC.EDU:mmt@ben.dciem.dnd.ca>
- Message-ID: <9301060350.AA00852@ben.dciem.dnd.ca>
- Newsgroups: bit.listserv.csg-l
- Date: Tue, 5 Jan 1993 22:50:51 EST
- Sender: "Control Systems Group Network (CSGnet)" <CSG-L@UIUCVMD.BITNET>
- From: Martin Taylor <mmt@BEN.DCIEM.DND.CA>
- Subject: Re: Why adopt a different theory?
- Lines: 41
-
- [Martin Taylor 930105 22:40]
- (Rick Marken 920105.1800)
-
- >Just today I was asked to help with a task at work that
- >involved trying to find a "reliable and valid test instrument" that could
- >be used to ascess some behavioral variables. Well, there is a problem;
- >how can PCT help solve it? First I could point out that the
- >tests are useless unless the reliabilities are on the order of .99. There
- >are no such tests and right off hand I don't know how to design them.
-
- It seems to me that you (and Gary Cziko some weeks ago) are confounding two
- issues: (1) what is going on in the tested person, and (2) what is the
- tester controlling for. Both of you in different ways argue that the
- statistical approach in the test of reliabilities less than 0.99 is
- pointless. Maybe it is, when you are dealing with the tested person
- (though I hold my options open to disagree later). But when you are
- dealing with the tester, the controlled variable may be something like
- "how many people exceed a criterion of success" in education or handling
- a complex instrument, or whatever. Generally speaking, the reference
- level for this variable might be 100%, so anything that increases the
- proportion is an action that forms part of the tester's functioning
- control system. They usually don't want to reduce the proportion of
- success if it is, say 90% and they were hoping for 80%, so it is only
- a one-sided control most of the time.
-
- The tester may have a conflicted control system (and often does). To get
- 100% passes {ay be possible, but only a great cost, and the tester also
- has a reference level that the cost be zero. So the two tester control
- systems conflict, resulting in less than 100% pass rate of the testees
- and greater than zero cost for whatever they do to the testees to improve
- the pass rate.
-
- All of what goes on in the tester relates to the ACTIONS of the testees,
- not the control systems of the testees, and it seems to me that the
- testers are justified in being happy that (more people survive the cancer
- longer| more students can read the instructions in a manual| more people
- are happy using the isntrument), even if the differences are measurable
- only statistically. It's a question of whose control systems are you
- worrying about, the tester's or the testees'.
-
- Martin
-