Abstract:
|
Cognitive diagnosis models(CDM) are partially ordered latent class models and are used to classify students into skill mastery profiles. Application of CDMs requires content expert knowledge of a Q matrix, which maps the test item to its corresponding required attributes or skills. Misspecification of the Q matrix has been shown to yield biased classification. We propose Bayesian frameworks for two popular CDMs (DINA model and rRUM) and estimate the corresponding Q matrix through Metropolis and Gibbs samplers. The proposed algorithms build upon prior research (Chen, Liu, Xu, & Ying, 2015) and ensure that the estimated Q matrices always satisfy the identifiability constraints. Monte Carlo simulation evidence is presented to support the accuracy of parameter recovery. We apply our algorithms to Tatsuoka's fraction-subtraction dataset.
|