- MCMC methods for item response theory (IRT) model estimation
- Introduction of informative priors in IRT models
- IRT model fit via posterior predictive model checking
- Statistical models for educational assessment
- Computerized adaptive testing (CAT)
- Automated test assembly (ATA)
- Inequalities in education and digital skills
The research activity deals with latent variable models and particularly with item response theory (IRT). Methodological contributions rely on Bayesian solutions based on Markov chain Monte Carlo (MCMC), i.e. the Gibbs sampler as estimation method and the posterior predictive model checks for model fit. The main contributions are: the integration of informative priors in the estimation of unidimensional models, the investigation of multidimensional approaches, and the goodness of fit for multidimensional models with complex structure. Applications deal with educational data (student assessment, INVALSI test, OECD PISA), psychometric data (intelligence tests, HADS scale), tourism data (residents' perceptions among the tourism industry), and digital skills (IEA ICILS), with a specific focus on inequalities.
