home *** CD-ROM | disk | FTP | other *** search
- Comments: Gated by NETNEWS@AUVM.AMERICAN.EDU
- Path: sparky!uunet!paladin.american.edu!auvm!DUKEMVS.BITNET!DOKIEM
- Message-ID: <STAT-L%92090309114626@VM1.MCGILL.CA>
- Newsgroups: bit.listserv.stat-l
- Date: Thu, 3 Sep 1992 09:07:00 EDT
- Sender: "STATISTICAL CONSULTING" <STAT-L@MCGILL1.BITNET>
- From: DOKIEM@DUKEMVS.BITNET
- Subject: ordinal or interval
- Lines: 34
-
- The number of possible alternatives that exist for handling
- ordinal data are numerous. They range from statistical approaches
- such as methods for categorical data to measurement approaches.
- The choice of solution depends primarily upon what you believe
- about the status of your data.
- If you regard your data as imperfect projections of an underlying
- dimension, then a measurement approach is called for. This can range
- from optimal scaling to fit a variety of statistical models to an
- analysis of the characteristics of the measurement procedure itself.
- The latter includes such ideas as item response theory (IRT) and
- mixed-variable factor analysis. The basic idea of item response
- theory is that the imperfect categorical responses are merely
- indicators of some underlying latent variable. Through the
- specification of the proper item-latent variable function or regression,
- estimates of the underlying variable can be obtained. The regression
- function generally includes parameters that may affect the reported
- scores as transformed from the underlying construct. In a quick but
- necessarily incomplete description, it is an attempt to understand the
- factors that affect measurement, and control for them. The
- result is an estimate of a continuous variable which not only fits
- a grater variety of statistical models, but which is more informative
- as to subtle individual differences. The classic text is Lord (1980),
- Applications of item response theory ...
- Of course, before we should take this to seriously, you should read
- Lord's entertaining essay in Psychological Bulletin some years back
- on "The statistical treatment of football numbers".
-
- Robert Terry
- Asst. Professor
- Dept. of Psychology
- Social and Health Sciences
- Duke University
- Durham, NC 27708
- DOKIEM@DUKEMVS
-