An IRT-MIMIC model for the analysis of university student careers

The paper focuses on the analysis of the performance of university students, with reference to first year compulsory courses. The main goal is to compare the exams in terms of difficulty, discrimination and use of the grades. Moreover, the paper aims at assessing how student careers depend on student and course characteristics. The analysis exploits an Item Response Theory approach where exams are treated as items, with a 2-Parameter Logistic model for the probability to pass the exams and a Graded Response Model for the ordinal items representing grades of passed exams. Course characteristics, such as the average student rating on teacher’s clarity, directly affect the items, whereas student characteristics, such as the type of high school, indirectly affect the items via the latent ability, even if some direct effects are allowed by fitting a MIMIC model with Differential Item Functioning. The analysis shows that IRT-MIMIC modelling is a flexible and powerful tool giving insights into the peculiarities of the exams and the role of course and student characteristics.