Logistic regression and its use in detecting nonuniform differential item functioning in polytomous items

Abstract

A computer simulation study was conducted to determine the feasibility of using logistic regression procedures to detect nonuniform differential item functioning (DIF) in polytomous items. A simulated test of 25 items was generated, of which the 25th item contained nonuniform DIF. The degree of nonuniform DIF in the 25th item was varied in four ways. Item scores were generated using Muraki's generalized partial credit model and the data were artificially dichotomized in three different ways for the logistic regression procedure. The results indicate that logistic regression is a viable procedure in the detection of most forms of nonuniform DIF; however, it was not sensitive to DIF that is uniform within score categories and nonuniform across score categories. Logistic regression procedures were also quite awkward in the polytomous case, because several regressions must be run per polytomous item and it was difficult to determine an omnibus result in most cases. Some logistic regression procedures, however, may be useful in the post hoc analysis of DIF in polytomous items.