Dear forum members,

I used WINSTEPS for purposes of estimating DIF contrasts for items from an achievement test (CONTEXT: I compared two groups of students; mean ability of students from group A was much lower than the mean ability of students from group B).

QUESTION:

For item X, the obtained DIF contrast amounted to 1 logit (and favored students from group A).
Can we interpret the described finding as follows:

If we compare equally able students from group A and group B, then the odds of solving correctly the item X is 2.7 times higher for students from group A than for correspondent students from group B.

OR (the same):

The odds of solving correctly item X, is 2.7 times higher for students from group A than for students from group B (provided that we control for ability).

I would highly appreciate any help.

Kulin

Yes, that looks correct, Kulin.

Most audiences have trouble with this type of arithmetic, so it may be easier to say something like:

if a student from group B has a 50% probability of success, then an equally able student from group A will have a 73% probability fo success.

Thank you very much, Mike!

Hi Mike... I'm trying to do the same than Kulin, however in my case I'm trying to compare differences between item location after item-linking. How can I write in terms of probability a difference of 0.44 logits, when group A item location :0.72 logits and groupB i loc: 0.28.

Is... Group A found 61% more complexity compared to the 61% of the groupB.
I reached this by: 0.72-0.28=0.44 (or 61%prob)

Thanks!!

Luis  

Luis, imagine a person of overall ability 0.28 logits in each group. Then, assuming the items are dichotomies, the person in group B has a 50% chance of success (0.28 ability - 0.28 difficulty = 0 logits). The person in group A has a 39% chance of success (0.28 ability - 0.72 difficulty = -0.44 logits).

You may want to choose a person with ability at the overall average of everyone as the reference person, instead of 0.28 logits.