DIF detection using logistic discriminant analysis and polytomous logisticregression

This study focused on the effectiveness in polytomous item DIF detection us ing Logistic Discriminant Analysis and Polytomous Logistic Regression. The simulated test consisted of 30 items with five categories per item, where D IF was manipulated in six items of the test. The conditions under study wer e: DIF effect size (0.5, 1.0 and 1.5), sample size (500 and 1000) and DIF t ype (nonuniform). The results suggest that Logistic Discriminant Analysis i s more accurate in DIF detection than Polytomous Logistic Regression. Howev er, the false positives rates for both procedures were similar.