In his talk, Professor Kaplan presented recent methodological developments in evaluating statistical models. Against the background of Bayesian approaches, possibilities were discussed of how to evaluate assertions and prognoses from inferential statistics in a comparative way. The chosen approach of Bayesian model averaging shows promising characteristics and allows a conceptually stringent consideration for the model’s uncertainty. The summation of multiple models into a single prognosis leads to an increased quality of predictions compared to prognoses based on single models. Corresponding effects were shown in the analysis of data from large-scale educational studies.