Item Response Theory

Abstract

Item response theory (IRT) seeks to model the way in which latent psychological constructs manifest themselves in terms of observable item responses; this information is useful when developing, evaluating, and scoring tests. After providing an overview of the most popular IRT models (i.e., those applicable to dichotomously keyed items) and contrasting them with the techniques used in classical test theory (CTT), the authors illustrate the application of IRT using data from the recently revised Myers-Briggs Type Indicator. These results highlight a number of IRT’s advantages, including (a) detailed descriptions of the performance of individual items, (b) indices of item- and scale-level precision that are free to vary across the full range of possible scores, (c) assessments of item- and test-level bias with respect to demographic subgroups, (d) measures of response-profile quality, and (e) computer-adaptive testing, which can dramatically reduce testing time.