The present study examines the problem of the development of psychometric models for multi-level designs, that aim to compare the medium level of subjects from different groups organized in levels hierarchically defined. A psychometric multilevel model based on the Item Response Theory (IRT) and a Bayesian procedure to obtain estimations in hierarchical models of IRT are presented. The model refers to dichotomous data and a one-dimensional latent trait, and put emphasis on the hierarchical aspect of the analysis. In addition to formally introducing the model, an illustration of the application of the procedure is presented by an example that includes empirical data referred to a test of mathematical knowledge that was applied to a sample of 1,000 Spanish students organized in schools from three different regions. The results provide information about each student, school, and region. Additionally, the syntax code used in the Bayesian estimation with the OpenBUGS and Stan programs is included in order to provide the reader with a tool that can be adjusted to his/her own research problem. Finally, the implications of the use of multilevel models and future research directions are discussed.
Revuelta, J., & Ximénez, C. (2014). Bayesian estimation of a multilevel psychometric model with random effects. Revista De Psicología, 23(1), 53–70. https://doi.org/10.5354/0719-0581.2014.32874
Twitter