No abstract

Examination of a test's internal structure can be used to identify what domains or dimensions are being measured, identify relationships between the dimensions, provide evidence for hypothesized multidimensionality and test score interpretations, and identify construct-irrelevant variance. The purpose of this research is to provide a methodological comparison of three methods for assessing the dimensionality and factor structure for binary scored exams in which guessing may be operating. The factor analytic methods in Mplus, NOHARM, and TESTFACT are compared in the context of analyzing the Multistate Bar Examination, a multiple-choice exam reflecting a number of complex integrated content domains. Results indicate that the methods support similar factor solutions. However, support for solutions with more factors and more indicators per factor is only obtained when guessing is modeled. The implications for evaluating the internal structure of multiple-choice tests are discussed.