[Rasch] Re: [MBC-Rasch] Rasch Digest, Vol 37, Issue 26
juho.looveer at gmail.com
Thu Apr 30 17:26:33 EST 2009
I don't know about using Rasch for experiments.
I guess, if you knew that two groups performed differently in, for example,
numeracy, and you then gave the group with lower performance some extra
help, you could retest both groups, expecting to see less dif. However, in
a total support program you would look for improvement overall. Perhaps dif
could help identify specific item types or areas of curriculum (which these
items address) where support had worked. however, you would need to be
conscious of test-retest and familiarity witht the relevant topics.
The definitions you quote from Broderson et al are right; however, I query
your comment that " the presence of DIF as non-uniform and its absence as
uniform"; I would not agree with that.
>From my experience, dif can be manifest in an item as either uniform or
non-uniform; That is why I find it interesting to use RUMM and look at dif
graphs for an item, rather than looking at one numerical indicator of dif.
It may be more difficult to interpret and understand, but it gives you more
And the presence of dif does not infer that there is something wrong - it
may just be that males are stronger and females are wittier (;-)
Dr Juho Looveer
On 4/30/09, Vahid Aryadoust(ELL) <ng0666907m at stdmail.nie.edu.sg> wrote:
> Thanks for your comment. You have implied that the Rasch is also suitable
> for experiments (It is my understanding?). For example, a treatment is
> administered to a groups and not another. Then, a test is given to them. To
> have some evidence for the success of the treatment, DIF should be observed
> once we divide the testees based on their group (experiment and
> non-experiment). Is it true?
> I also found interesting ideas in John Brodersen et al.'s article.
> I found a classification of DIF (uniform and non-uniform) "An item that
> functions with the same constant magnitude of difference across a construct
> measured possesses uniform DIF. In contrast, non-uniform DIF is
> characterised by an uneven difference in item function across the latent
> variable measured."
> I have always thought that DIF is of a yes-no nature. It either exists or
> does not. I find that John and his colleagues define the presence of DIF as
> non-uniform and its absence as uniform. I am not really sure whether I am
> right. I appreciate comments and guidance from anyone.
> S. Vahid Aryadoust
> English Language and Literature Department
> National Institute of Education
> Nanyang Technological University
> + 65 67903578
> Quoting Juho Looveer <juho.looveer at gmail.com>:
>> If the two groups ARE different on some factor (e.g. you say females
>> are more educated), then there are various scenarios . . .
>> 1. The factor where there is a difference (amount of education) does
>> impact on the skills/traits being assessed - then I would expect to
>> see evidence of dif.
>> 2. The factor does NOT impact on the skills/traits being assessed;
>> then you would not expect to see any evidence of dif; if you did see
>> dif, then it would be due to some other cause.
>> How to differentiate the cause of dif? that's another issue
>> 3. The factor DOES impact on the skills/traits being assessed, but
>> acts to make up for some inherent difference e.g. if you had a group
>> of females who had been doing specific strength training, then their
>> performance in a strength test might be equal to the performance of
>> the males, assuming that males were generally stronger in the first
>> Dr Juho Looveer
>> On 4/29/09, rasch-request at wu-wien.ac.at <rasch-request at wu-wien.ac.at>
>>> Send Rasch mailing list submissions to
>>> rasch at wu-wien.ac.at
>>> To subscribe or unsubscribe via the World Wide Web, visit
>>> or, via email, send a message with subject or body 'help' to
>>> rasch-request at wu-wien.ac.at
>>> You can reach the person managing the list at
>>> rasch-owner at wu-wien.ac.at
>>> When replying, please edit your Subject line so it is more specific
>>> than "Re: Contents of Rasch digest..."
>>> Today's Topics:
>>> 1. Re: Interpretation of DIF (John Brodersen)
>>> Message: 1
>>> Date: Wed, 29 Apr 2009 12:02:31 +0200
>>> From: John Brodersen <J.Brodersen at gpmed.ku.dk>
>>> Subject: Re: [MBC-Rasch] Interpretation of DIF
>>> To: The Matilda Bay Club <rasch at wu-wien.ac.at>
>>> Message-ID: <49F825B7.4080404 at gpmed.ku.dk>
>>> Content-Type: text/plain; charset="iso-8859-1"
>>> Dear Vahid,
>>> you are right. You shouldn't identify DIF no matter the level of
>>> education. Together with some colleagues, mainly professor Svend
>>> Kreiner, I have written a paper about DIF using Rasch modelling and how
>>> to adjust for uniform-DIF. I have attached the paper and hope it can
>>> help you in your further work.
>>> Best regards
>>> Vahid Aryadoust(ELL) skrev:
>>>> Hi all,
>>>> I have a quick question about DIF and its interpretation. I have given
>>>> a 40-item test of listening comprehension to a group of participants.
>>>> I used DIF to find out if some items were biased towards some
>>>> subgroups. Female participants are more educated than males. Now I
>>>> find DIF for gender in some items. My question is "Am I right to think
>>>> even if a group of participants is more educated than another, there
>>>> should not be any DIF unless the question is biased toward a group?"
>>>> In other words, this is not the background that should affect test
>>>> performance, even if this background is students' education; am I on
>>>> the right track?
>>>> I appreciate your assistance.
>>>> S. Vahid Aryadoust
>>>> English Language and Literature Department
>>>> National Institute of Education
>>>> Nanyang Technological University
>>>> + 65 67903578
>>>> Rasch maillist - Rasch at wu-wien.ac.at
>>> John Brodersen
>>> Speciall?ge i almen medicin, ph.d., lektor
>>> K?benhavns Universitet
>>> Afdeling for Almen Medicin
>>> Institut for Folkesundhedsvidenskab
>>> Center for Sundhed og Samfund
>>> ?ster Farimagsgade 5, bygning 24Q, Postboks 2099
>>> 1014 K?benhavn K
>>> TEL 3532 7960
>>> DIR 3532 7592
>>> FAX 3532 7946
>>> MOB 2027 6181
>>> Kontor 24.1.21
>>> j.brodersen at gpmed.ku.dk
>>> Logo UK
>>> -------------- next part --------------
>>> A non-text attachment was scrubbed...
>>> Name: DIF_paper_published_version.pdf
>>> Type: application/pdf
>>> Size: 707980 bytes
>>> Desc: not available
>>> Url :
>>> Rasch maillist - Rasch at wu-wien.ac.at
>>> End of Rasch Digest, Vol 37, Issue 26
>> Rasch maillist - Rasch at wu-wien.ac.at
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Rasch