[Rasch] DIF across gender
Fabio La Porta
fabiolaporta at mail.com
Wed Mar 8 22:38:58 AEDT 2017
the Rasch model tells us that in order to achieve measurement, items should
satisfy the homogeneity requirement, that is their stochastic ordering
should be invariant (i.e. should remain the same regardless of the the
ability of the subjects). This requirement cannot not be merely assumed but
needs to ne tested.
At the same time, the Rasch model assumes that the homogeneity requirment
needs to satisfied also across subgroups of patients with respect to
relevant exogenous variables such as gender, age, etc. Again, the
satisfaction of this requirement (absence of DIF) cannot be merely assumed
but needs to be demonstrated, otherwise we will introduce in the
measurement process an important bias.
So, I would respond to the referee by turning his/her question the other
way around, that is we test for DIF by gender in order to confirm that
there is no item bias by this relevant demographic factor.
Having said that, if an item displays DIF, it will be your job to find a
substantive explanation for this finding, bearing in mind that testing for
DIF, as any other statistical test, may be affected by type I error.
I hope this helps.
2017-03-08 12:22 GMT+01:00 Amir Pakpour <amir.pakpour at gmail.com>:
> Dear All
> I have done a Rasch model on a self reported measure in cardiac
> patients. i examined DIF for each item across gender (male vs.
> female) . However, the reviewer wrote me that " I doubt the reasoning
> around the DIF. For me it is not clear why we would think that based
> on gender would be different?"
> I think that Measurement invariance tests the level of consistency in
> the understanding and interpretation of a measure by persons that
> differ in sociodemographic characteristics (for example by gender)
> and/or are from different cultural, ethnic, and linguistic
> backgrounds. Why we should have a special justification for performing
> the DIF?
> any suggestions are welcome
> All my best
> Amir H Pakpour, Ph.D.
> Associate Professor,
> Department of Public Health,
> Qazvin University of Medical Sciences,
> Qazvin, Iran.
> Rasch mailing list
> email: Rasch at acer.edu.au
> web: https://mailinglist.acer.edu.au/mailman/options/rasch/
Powered by Gmelius
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Rasch