Robustness of person-fit decisions in computerized adaptive testing

Person-fit statistics test whether or not the likelihood of a respondent’s complete vector of item scores on a test is low given the hypothesized item response theory (IRT) model. This binary information may be insufficient for diagnosing the cause of a misfitting item-score vector. This paper applies different types of person-fit analysis in a computer adaptive testing context and investigates the robustness of several methods to multidimensional test data. Both global person-fit statistics to make the binary decision about fit or misfit of a person’s item-score vector and local checks are applied. Results showed that there are differences between the methods with respect to the robustness in a multidimensional context and that some methods are more useful than othermethods.