[Rasch] Fitting the RM

Thomas Salzberger Thomas.Salzberger at wu-wien.ac.at
Tue Jul 8 00:06:10 EST 2008


I don't add anything new to what Trevor already said but just want to 
reinforce the idea.
Retrospectively, it was one of the most important insight I gained 
concering measurement from a Rasch point of view.
We do not (only) choose the Rasch model because of its elegance or 
simplicity (relative to more complex IRT models) but because of its 
unique properties and its relationship to the principles of 
fundamental measurement. That's why the model enjoys a special 
status, it has priority over the data insofar as we are not willing 
to abandon or change the model if it happens to misfit data. Instead, 
we say that the data misfit the model, in other words the data don't 
have the structure that allows us to derive linear measures.

Personally, I would restrict the usage of the term "to fit" to 
statements like "data fit the model" or "data don't fit the model". I 
would rather not use it in the context of an activity. I would 
certainly not say "I fit the model to the data" as this implies 
changes to the model which are unjustified. Likewise "I fit the data 
to the model" may suggest changes to the data (although in some cases 
we do change the data in a sense, e.g. when rescoring items, but then 
this should not be the final conclusion).

In any case, I would distinguish between the Rasch model (with all 
its philosophical underpinnings) and an 1pl-model (without these 
theoretical foundations) in a more general IRT context. Most IRT 
people do appreciate the Rasch model's elegancy but they don't 
approve its special relationship to measurement. For them it is just 
convenient and if it misfits they switch to a more general model and 
fit that model to the data. (In most cases they don't even start with 
the Rasch model as they see it as overly restrictive.)

Best wishes

At 12:18 07.07.2008, Trevor Bond wrote:
>Dear Anthony,
>We fit data to the model: see B&F2, chap 12 for the detail.
>Many others try to find models to fit the data in order to describe the data
>we aim to have the data to fit the model to create measures...or to 
>understand 'why not?'
>At 2:57 AM -0700 7/7/08, Anthony James wrote:
>>Dear all,
>>I have come across expressions such as "fitting the Rasch model (or 
>>2PL, 3PL) to Test X" or "the Rasch model fits (or does not fit)", 
>>"the Rasch model was fitted".
>>My question is when we "fit the Rasch model to a test" we are more 
>>emphasizing the model rather than the data. I was curious to know 
>>if these wordings have any special meaning and implications. Or do 
>>they simply mean Rasch analyzing the data and finding the 
>>misfitting items and persons and other routine analyses?
>>Rasch mailing list
>>Rasch at acer.edu.au
>Trevor G BOND Ph D
>Professor and Head of Dept
>Educational Psychology, Counselling & Learning Needs
>D2-2F-01A EPCL Dept.
>Hong Kong Institute of Education
>10 Lo Ping Rd, Tai Po
>New Territories HONG KONG
>Voice: (852) 2948 8473
>Fax:  (852) 2948 7983
>Rasch mailing list
>Rasch at acer.edu.au
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://mailinglist.acer.edu.au/pipermail/rasch/attachments/20080707/e151a5ca/attachment.html 

More information about the Rasch mailing list