Large-scale assessment in the education field is key in every Country. In Italy, the institute that is in charge of evaluating pupils’ proficiency is the INVALSI, via a set of standardized tests, that go in parallel with traditional school evaluation. Data collected in such a way at the individual level pose a statistical challenge, given the nested structure of students-classroom-school and the repeated measure longitudinal observations that are obtained for each student who performs the tests in several years during their scholastic career. We propose in this context the streamlined version of the mean-field variational Bayes (MFVB) algorithm for linear mixed models with crossed random effects in order to obtain plausible predictors of pupils’ performances. The results and interpretation of model coefficients are in line with the literature on educational data.
