Robert Henson
Department of Educational Research Methodology
University of North Carolina at Greensboro
Markov Chain Monte Carlo Estimation for a Family of Cognitive Diagnosis
Models
Cognitive diagnosis models are a set of constrained latent class models in
which classes are determined by mastery or non-mastery of a set of
dichotomous latent variables, or skills. Because of the possibility of
additional information that may improve learning, cognitive diagnosis models
hold great potential. Specifically, by knowing each examinee's skills,
tailored lesson plans can be used, which can make teaching more effective.
This talk first provides a general overview of cognitive diagnosis models.
Then, using a log-linear model for latent variables (Hagenaars, 1993), a
family of cognitive diagnosis models is defined and the relationships
between many common models are analytically expressed. Finally, the
estimation of this model using Markov chain Monte Carlo is discussed and a
small example is provided.
Back to Colloquium Series