The dangers of statistical tests of nuclear structure models

It is common to assess the quality of nuclear models using statistical tests such as chi-square measures. Yet such tests, carried out without special care, can be extremely misleading. These dangers center on three issues: how to deal with data of differing precision, which observables to focus on, and theoretical uncertainties. These topics will be discussed with simple (idealized) examples. Time permitting, an application to a model of a partial dynamical symmetry based on SU(3), will be discussed, that illustrates the pros and cons of this model, the importance of valence nucleon number, and the issues of testing models mentioned above. This document is composed of an abstract and the slides of the presentation. (author)