Diagnostic Classification Models

Diagnostic classification models, also called cognitive diagnosis models, are a class of psychometric models developed to classify test-takers into discrete categories of mastery or non-mastery on a defined set of specific skills or attributes, rather than producing a single continuous ability score of the kind produced by classical test theory or item response theory. Each test item is mapped, in what is commonly called a Q-matrix, to the particular skill or combination of skills it requires, and a test-taker's pattern of correct and incorrect responses is used to estimate which specific skills that person has and has not mastered. Foundational work on this approach includes the rule space methodology introduced by the psychometrician Kikumi Tatsuoka in the early 1980s, and later formal models such as the deterministic input, noisy and gate model, commonly known as the DINA model, which further developed the statistical machinery for estimating skill-mastery profiles from test data. Because they report mastery status on individual skills rather than a single overall score, diagnostic classification models are used especially in formative and diagnostic testing, where the goal is to give teachers and students fine-grained feedback on precisely which skills still need instruction, rather than only a summary judgment of overall performance.