[Rasch] Two Groups, Pre & Post

Petroski, Greg PetroskiG at health.missouri.edu
Sat May 6 04:11:05 EST 2006

I hope I didn't sound glib.  I assumed that on this list I was preaching
to the choir.  However, the "obvious reason(s)" for me is that ordinal
scales do not have meaningful units of measurement.  It is true that we
are all comfortable with reporting "an mean change of k points" and tend
to forget that "point" is meaningless.  In my consulting role I have
often asked a client "what kind of change is clinically meaningful?".
This question most often comes up in sample size planning.  I would
guess that 4 out of 5 times even the users of ordinal scaled "measures"
can not articulate what a meaningful change is.  At some point it is big
enough to care about, and at the other extreme it is small enough to be
irrelevant, and there is a huge grey area in between.

To the typical investigator the transformation to "logits" does nothing
to enhance their understanding of the results.  Frankly, I don't like
the logit scale either, it really is incomprehensive to most people.
But there is a unit of measurement and so it seems better, when
possible, to use a measure rather than a set of points. 

Additional comments are welcome as to how to best present results using
Rasch scoring in an applied manuscript -- i.e. where methods are not
going to get a lot of space. 

Greg Petroski, Ph.D.
Health Management and Informatics &
Office of Medical Research/Biostatistics
137 Hadley Hall (DC 018)
University of Missouri - Columbia,
Columbia, Mo.  65212
ph:  (573)882-3373
fax: (573)882-7909

-----Original Message-----
From: Nianbo Dong [mailto:dong.60 at gmail.com] 
Sent: Thursday, May 04, 2006 8:15 PM
To: Petroski, Greg; rasch at acer.edu.au
Cc: dong.60
Subject: Re: [Rasch] Two Groups, Pre & Post

Hi Greg,

I have a question for you.
To my knowledge, we can use Graded Response Model (GRM) or Partial
Credit Model (PCM) to calibrate ordinal scaled data.
GRM is an indirect IRT model, which includeds 2-step process.
PCM is generaized from 1PL, i.e. Rasch model (however, PARSCALE uses
To decide which model we should use, we need to compare these models
based on -2 log likelihood and item fit statistics to see which model
fit the data better. My question is what the reasons are to "apply Rasch
scoring for the obvious reasons". I'm sorry that I don't know what
"Rasch scoring" 
means here.

Thank you.

Nianbo Dong

First year Ph.D. student
Policy Research, Evaluation, and Measurement (PREM) Graduate School of
Education University of Pennsylvania

More information about the Rasch mailing list