[Rasch] Post-hoc adjustment for interactions in Facets
rmt at rasch.org
Sun Aug 5 07:13:22 EST 2012
It sounds like you are testing the null hypothesis "no raters are biased
relative to item type". This would require a Bonferroni (or similar)
adjustment to individual-rater t-tests, or the use of a global chi-square
(or similar) test.
At 8/4/2012, you wrote:
>I have been looking at some interactions in Facets, and was recently
>challenged about the use of the t-statistics to show significance. If I'm
>using the t-statistics from the bias/interaction report, is it appropriate
>to do a post-hoc adjustment to avoid Type II error?
>My analysis was testing the null hypothesis that rater severity does not
>vary over two types of items. A significant interaction would seem to
>suggest that rater interpretation of the construct is not invariant across
>Thanks for your help!
More information about the Rasch