[Rasch] DIF test with 2-item scale

Bond, Trevor trevor.bond at jcu.edu.au
Wed Jan 4 09:46:26 AEDT 2017


Dear Svend,
Is DIF 'evidence against the Rasch model' or evidence that the data do no conform to its requirements for measurement? Perhaps a tendentious question, but I seek clarity for me, and perhaps, others who lurk on the forum.
Collegially
Trevor

Prof Trevor G BOND

> On 4 Jan 2017, at 2:12 am, Svend Kreiner <svkr at sund.ku.dk> wrote:
> 
> You can (and should) test for dif. If evidence of dif turns up, it is evidence against the Rasch model. The problem, as Thomas has already said, is that you have no way to determine whether one of the items have no dif.
> 
> Also, you would have to have many more items fitting a Rasch model to convince anyone that you have a valid measure of anything. Three items are better. In such cases you can meaningly do a statistical analysis of the fit to the Rasch model and the claims of both no dif and local independence, but what about the non-statistical requirement of content validity? That would require many more items, I think.
> 
> Svend
> ________________________________________
> From: Rasch [rasch-bounces at acer.edu.au] on behalf of Randy and Shelley MacIntosh [srmac at fastmail.com]
> Sent: Monday, January 02, 2017 8:05 PM
> To: rasch at acer.edu.au
> Subject: [Rasch] DIF test with 2-item scale
> 
> I have several two-item scales. I am guessing that it is impossible to
> test for DIF.
> Is that correct?
> Any suggested reading on the minimum number of items required to test
> for DIF in a Rasch model would be appreciated.
> 
> Thanks,
> Randy MacIntosh
> 
> ________________________________________
> Rasch mailing list
> email: Rasch at acer.edu.au
> web: https://mailinglist.acer.edu.au/mailman/options/rasch/svkr%40sund.ku.dk
> ________________________________________
> Rasch mailing list
> email: Rasch at acer.edu.au
> web: https://mailinglist.acer.edu.au/mailman/options/rasch/trevor.bond%40jcu.edu.au


More information about the Rasch mailing list