A0831
Title: Variational Bayesian estimation in diagnostic classification models
Authors: Keiichiro Hijikata - The University of Tokyo (Japan)
Motonori Oka - London School of Economics (United Kingdom)
Kazuhiro Yamaguchi - University of Tsukuba (Japan)
Kensuke Okada - The University of Tokyo (Japan) [presenting]
Abstract: Recent developments in variational Bayesian estimation methods for diagnostic classification models (DCMs) are discussed. The widespread use of information technologies in educational environments has created a demand for cognitive diagnosis and personalized feedback based on data obtained from learning systems. However, data from these systems often include a large number of respondents, items, and attributes. To efficiently conduct cognitive diagnosis in such situations, variational Bayesian estimation methods have been developed for a large class of diagnostic classification models, including the deterministic input, noisy "and" gate (DINA) model, deterministic input, noisy "or" gate (DINO) model, multiple-choice DINA model, and saturated DCM. The proposed algorithm consists of iteratively repeating two steps, the variational E-step and the variational M-step, until convergence. This algorithm is an important component in the scalable estimation of the Q-matrix in the DINA model. To facilitate the application of the proposed methods and further methodological developments, an R package, variationalDCM, has been developed that implements the proposed estimation methods for DCMs. The developed algorithm and implementation provide an effective framework for the study and application of DCMs in modern educational environments.
