Dear Rasch-stars,

I performed an interaction analysis using Facets and I am interested in determining whether rater severity interacts with item type. Based on my understanding of the interactions in Facets, the t-statistics are testing the null hypothesis that a rater's severity on a certain type of item is not significantly different than the overall rater severity on that type of item.

However,when I'm looking at t-tests for 15 raters on these two types of items, it seems like it might be appropriate to do a post-hoc adjustment for the t-statistics in order to avoid Type II error. Would a Bonferroni adjustment be appropriate?

Thanks!
Stefanie

Stefanie: a Bonferroni adjustment is appropriate if your hypothesis is "These raters are unbiased." No adjustment is needed if your hypothesis is "This rater is unbiased." And, of course, we choose the hypothesis before looking at the results of the statistical tests.