Journal of Educational Measurement Psychology (miscellaneous)
Population Difficulty and Discrimination in Item Response Theory: A Complement to Traditional Item Parameters
Abstract
Abstract Traditional Item Response Theory (IRT) parameters for item difficulty and discrimination often depend heavily on the specific assumptions and constraints of the chosen model. This study introduces and evaluates population difficulty and population discrimination as complementary metrics that generalize across dichotomous and polytomous item types and apply to both unidimensional and multidimensional models. These metrics can be used independently or as a complement to traditional parameters. They also readily extend to semiparametric or nonparametric modeling approaches. We compare the population‐based metrics to traditional item parameters through a series of simulations and an empirical example. Our findings suggest that population‐based metrics are robust metrics for evaluating item characteristics across a variety of different populations or IRT models.