Analyzing Knowledge Structure: an Application of Graphical Models to a Medical Licensure Exam

Supporting student success well in a global economy under the very realworld constraints of scarce resources requires certain economies of instructional practice. Teachers with ever-more-limited time and ever-increasing demands need to make critical decisions concerning educational interventions for individual students quickly and accurately. Supporting student success thus entails supporting teachers by providing timely, high-quality, detailed information about students’ current progress. Most of today’s educational assessment tools, however, usually provide only high-level information about broad educational domains (such as mathematics, reading, science). Measurement models for rapid educational response need to provide analysis of student achievement on a variety of sub-domains, and must eventually do so with a modest number of items. Several classes of models are common choices for providing sub-domain ability estimation, including Multidimensional Item Response Theory (MIRT; Reckase, 2009) and diagnostic classification models (DCM; e.g. Rupp, Templin, and Henson, 2010). However, most of these models leave the structure of the underlying latent abilities unspecified, and it may be possible to achieve greater measurement precision by explicitly modeling the sub-domain relationships. Graphical models serve as one means for modeling these relationships.