Parsimonious matrix normal mixtures: an application to university students indicators

The analysis of the teaching efficiency plays a fundamental role for universities. This process can be measured by evaluating several aspects such as the successful completion of the studies by the students or the students/teacher’s ratios. Here, these two areas are investigated for the University of Catania by using two matrix variate datasets. Such data structure allows the simultaneously consideration of several variables measured over different years. To detect possible underlying group structures, and for mitigating overparameterizations issues, parsimonious matrix normal mixtures are fitted to the datasets. The results show the existence of different groups in the data, each having distinctive characteristics, providing than useful information for the university managers.