Posterior predictive model checking for cognitively-based
psychometric models
Roy Levy
Bayesian inference and cognitively-based statistical models are two emerging and
converging lines of research in psychometrics. Model criticism remains a relatively
unexplored aspect of Bayesian approaches to psychometric modeling. This
presentation will discuss posterior predictive model checking, a powerful approach to
model criticism, with an emphasis on cognitively-based models. Examples will be
drawn from both methodological and applied work. Finally, I will address the role of
posterior predictive model checking in the larger scope of the design, development,
and use of assessment systems that are grounded in the substantive analysis of the
domain.