[Rasch] unfolding models and Rasch

Stephen Humphry stephen.humphry at uwa.edu.au
Wed Aug 29 11:25:54 EST 2007

You ask whether there is some way to model this with a Rasch model.

If (i) statements are ordered with respect to a latent trait and (ii) the
respondents must choose one of a set of such ordered categories, then the
polytomous Rasch model can be applied. The probability of responding in a
category 1 and 7 are both low when a respondent is most likely to respond in
category 4 if the thresholds are ordered and reasonably spaced apart. The
expected score should rise monotonically, but the proability of a response
in each ordered category does not rise monotonically in the polytomous Rasch

I was involved in a study where this kind of design was used and the
scenarios were ethical dilemmas.

On the other hand if, for example, respondents must agree or disagree with
each statement as a seprate item, you would probably need to use an
unfolding model. Andrew Kyngdon has offered to help there.

Or do you have some other response format in mind?



Dr Stephen Humphry
Graduate School of Education
University of Western Australia
35 Stirling Highway
Mailbox M428
P: (08) 6488 7008
F: (08) 6488 1052

-----Original Message-----
From: rasch-bounces at acer.edu.au [mailto:rasch-bounces at acer.edu.au] On Behalf
Of Donald Bacon
Sent: Tuesday, 28 August 2007 9:25 PM
To: rasch at acer.edu.au
Subject: [Rasch] unfolding models and Rasch

Hi Raschers -

   I'm working with a scale wherein respondents read a scenario and then
rate how similar each of several statements are to their own way of thinking
about the scenario.  In this way we hope to locate where each respondent is
on a latent trait called reflective judgment, as well as calibrate each of
the statements.

   It occurs to me, however, that this is not like a common Rasch analysis.
For example, a respondent at a level 4 on a 7 level latent trait will rate
both a level 1 statement and a level 7 statement as being highly dissimilar
to his or her way of thinking.  Thus, response probabilities do not increase
monotonically with respondent ability.  Would some type of unfolding model
work better?  Or is there some way to model this in Rasch?


Thanks for any insight you might have -


Don Bacon

Associate Professor of Marketing

Daniels College of Business
University of Denver
dbacon at du.edu, 303-871-2707
Rasch mailing list
Rasch at acer.edu.au

More information about the Rasch mailing list