Methods for testing DIF in IRT are model based, in that they require fitting a latent variable model in both groups. IRT provides a convenient and powerful framework for evaluating whether an item is functioning differentially by comparing each group’s parameter estimates or IRF. In fact, DIF can be defined explicitly from the IRT models as the difference in the probability of responding to a category (e.g., correct response for a dichotomous item) for examinees with the same ability from different populations (Lord, 1980). This definition is consistent with comparing the IRFs or item parameter values between groups. In other words, if the parameter values or IRFs are identical in both populations, then the probabilities of responding to a category are the same and therefore the item is DIF-free. However, if the IRFs differ in the populations, then the item is functioning differentially (Lord, 1980).