[Rasch] Invariance

Thomas Salzberger Thomas.Salzberger at wu-wien.ac.at
Fri Jan 6 00:28:06 EST 2006

Zitat von Rama Guinda <guinda36 at hotmail.com>:


you can take your ability split as a person factor. Estimate the item measures
from the pooled data and do a test for differential item functioning (low
versus high ability) which esentially is a test of invariance.

Best wishes

> Dear All,
> I have a question. Thanks for any reaction to my problem.
> I Rasch analysed 4 polytomous items with the Winsteps. The items where 4 
> texts each with 20 gaps to be filled by the test-takers. In order to avoid 
> the problem of local dependence I considered each text a super item. I had 
> 160 cases.
> The item measures are .74, .06, -.29 and -.50.
> Then I divided the group into two (a high ability and a low ability group) 
> on the basis of a  reading test which contains 14 dichotomously scored 
> items. Students with scores from 0 to 7 went into the low ability group and 
> the ones with scores from 8 to 14 went into the high ability group. The 
> reading test has a significant but low correlation with the gap fill task 
> (.53).
> I had 98 cases in the low ability group and 62 cases in the high ability 
> group. To study the item invariance I separately analysed the data for each 
> group.
> The item measures for the low ability group are: .62, .10, -.25, -.47.
> The item measures for the high ability group are: .94, .02, -.38, -.58.
> The order of the difficulty of the items remains the same in all three 
> groups.
> Question 1: What is the test of invariance for the item measures? Have they 
> really remained invariant within the allowable range of error?
> Best
> Rama
> _________________________________________________________________
> Tuff to bluff! Get Abhishek before he gets you! Enter the contest 
> http://utv.movie-talkies.com/bluffmaster/contest.html
> _______________________________________________
> Rasch mailing list
> Rasch at acer.edu.au
> http://listserv3.acer.edu.au/mailman/listinfo/rasch

Liebe Gruesse / Kind regards
Thomas Salzberger

More information about the Rasch mailing list