Diagnostic Measurement in ERM
Diagnostic classification models (DCMs) are used to provide formative feedback to students on each of several attributes (knowledge or skills) represented on an assessment. Traditional measurement models have tended to focus on providing a single continuous score. DCMs, by contrast, provide classifications (e.g., mastery, non-mastery) on each attribute. The resulting mastery profiles provide test takers (or their parents and teachers) with information that can be used to target learning interventions where they are needed most. As an example, a traditional math test might return a single math score. A student may also receive normative information (i.e., a percentile rank against other test takers). A diagnostic classification model might return a profile that indicated whether a student had mastered multiplication, division, and orders of operation. Then, a student’s mastery profile could be used to suggest remediation only where needed.
Dr. John Willse and Dr. Bob Henson currently are leading student research groups on DCM topics. Recent work has focused on integrating several frameworks for examining the multidimensional structure of assessment data and for accomplishing the estimation of different diagnostic modeling approaches. These working groups are intended both to advance the literature on diagnostic models and to develop expertise in doctoral student researchers.
What does the figure at the top of the page have to do with any of this? That figure is a path diagram used to represent the relationship between observed variables (items 1 through j) and the latent diagnostic classifications (through ). Several types of relationship are shown in the diagram (from left to right): simple structure, conjunctive complex structure, and compensatory complex structure. Don’t worry if that is not clear. We have research groups that will help get you up to speed.