This article discusses the limits in applying mixed effects models in a routine evaluation of the efficacy of European funded professional education. In particular, we discuss the sensitivity of these models to changes in the underlying assumptions, usually related to the within-institution correlation, with reference to a data set on the evaluation of public-funded programs for professional education in Northern Italy. The models are evaluated with respect to the variable selection process, comparing the differences in the selected model (and consequently on the rank of the institutions that carry out the education courses) if the hypothesis on the random effects, on the observations correlation and on the variance function form are changed.
