A1538
Title: VariationalDCM 2.0: An updated R package for variational Bayesian inference in diagnostic classification models
Authors: Kensuke Okada - The University of Tokyo (Japan) [presenting]
Keiichiro Hijikata - The University of Tokyo (Japan)
Motonori Oka - London School of Economics (United Kingdom)
Kazuhiro Yamaguchi - University of Tsukuba (Japan)
Abstract: Variational Bayesian methods provide a computationally scalable alternative to traditional Markov chain Monte Carlo (MCMC) techniques, making them highly effective for large-scale data analyses often encountered in educational and psychological assessments. Diagnostic classification models (DCMs), a class of psychometric models, are employed to diagnose latent skills or attributes based on item response data, allowing for the assessment of specific skill mastery or non-mastery in respondents. The enhanced R package, variationalDCM, leverages these methods to efficiently and scalably estimate DCMs. It is available on CRAN, making it easily accessible for researchers. In the latest version, 2.0, several important updates have been incorporated. First, it integrates five DCMs into a unified fit function, streamlining model specification and improving usability. Second, the package now outputs results as an S3 class, enabling easy integration with other R packages and simplifying post-processing. Additionally, a newly implemented summary function offers quick access to key diagnostics and parameter estimates, enhancing the interpretability of results. The advantages of variational Bayesian inference are emphasized for DCMs, particularly its ability to deliver fast, approximate solutions with strong theoretical guarantees and demonstrate the functionality of the package through illustrative examples.
