[Rasch] current issues of research in psychometrics?

Andrew Kyngdon akyngdon at lexile.com
Thu May 3 16:44:19 EST 2012


Mike,



Psychometricians are metrologists. Is it the fault of the metrologists when
their customers use antiquated measuring instruments?



Funny you mention antiquated measuring instruments. I was just on a Skype
call with Steve Humphry and I put the following two points to him. Firstly,
that pencil and paper based tests are an old and inherently limited
observational methodology for individual differences in cognitive
abilities. In my opinion, there is just no getting around the problem that
such tests yield partial orders of test score response patterns. Even when
you have a theory to get around that problem, such as the Lexile Framework,
other problems exist which you trip you up. For example, putting in the
same test two imbedded sentence cloze reading items with the same stem, but
different cloze sentences and foils (violation of local independence is
highly probable). I really do think that psychometricians need to embrace
information technology, similarly to what Jack Stenner has done with the
Oasis platform and the Lexile Framework.



The second point is that there is absolutely nothing in the logic of the
Rasch model that says it must be only applied to the calibration of
psychometric tests. Steve and I are currently formulating a study using the
Poisson Counts Rasch Model which will not use data from psychometric tests
at all. Indeed, Rasch himself was first concerned with modelling reading
errors, not the calibration of tests.



The scientific study of individual differences in cognitive abilities needs
to move on from the pencil and paper based psychometric test, whatever
practical uses in education, licensure, etc it may have.



My summary of many Psychometrika papers is: a good idea,  followed by many
pages of obscure algebra, finished up with a trivial example.



I can’t argue with you there.



It is also an unfortunate accident of history that Lee Cronbach
misunderstood Georg Rasch when they met in 1955
www.rasch.org/rmt/rmt61c.htm. If Lee Cronbach had understood Georg
Rasch's methodology, we might now
know the Rasch model as Cronbach's Beta model!



Thanks for that. Yes, it strikes me how confronting some people feel that
there is actually a plausible zero point for cognitive abilities. A rock
has no reading ability. My 2 year old son recognised his written name
yesterday for the first time, so has more reading ability than the rock,
but less ability than you or me. A Supreme Court judge most likely has a
reading ability far greater than yours and mine. So the Poisson Counts
Rasch Model (or “multiplicative model”) and its implication of an absolute
zero point is not without plausibility, at least for reading errors.



Another example of a zero point is “no change from the status quo” or
“reference point” of modern utility theory. In cumulative prospect theory
and expected utility theory, the binary gamble ($100, 0; $0, 1) has zero
utility as there is zero probability of winning $100 and complete certainty
of winning nothing. Utility theorists are happy with this.



A human example of a true zero point is Julia Gillard, the Prime Minister
of Australia. She has proven to everyone that she has zero ability to
govern a country. J



Andrew





*From:* rasch-bounces at acer.edu.au [mailto:rasch-bounces at acer.edu.au] *On
Behalf Of *Mike Linacre
*Sent:* Thursday, 3 May 2012 12:17 PM
*To:* rasch at acer.edu.au
*Subject:* Re: [Rasch] current issues of research in psychometrics?



Folks:

Andrew K. wrote: "the most viewed paper online was Cronbach's original 1951
paper on "Alpha".

Psychometricians are metrologists. Is it the fault of the metrologists when
their customers use antiquated measuring instruments? Imagine if the most
viewed paper in a metrology Journal was about "knotted ropes", who would we
blame? Surely not the manufacturers of laser measuring devices.

Perhaps it is a problem with Psychometrika that it tends to make modern
measurement methods appear so obscure. My summary of many Psychometrika
papers is: a good idea,  followed by many pages of obscure algebra,
finished up with a trivial example.

It is also an unfortunate accident of history that Lee Cronbach
misunderstood Georg Rasch when they met in 1955
www.rasch.org/rmt/rmt61c.htm. If Lee Cronbach had understood Georg
Rasch's methodology, we might now
know the Rasch model as Cronbach's Beta model!

Food for thought ...

Mike L.

Mike Linacre
rmt at rasch.org www.rasch.org/rmt/ Latest RMT: 25:4 Spring 2012
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://mailinglist.acer.edu.au/pipermail/rasch/attachments/20120503/d522a79f/attachment.html 


More information about the Rasch mailing list