Model-Based Ranking of Schools

This report presents an empirical study of the problem of ranking schools in terms of their quality. The unadjusted ranks are compared with rankings obtained by various adjustment schemes based on ANCOVA and random effects models. The advantage of random coefficient models is highlighted by the reanalysis of a 1959 data set that contains information about 1290 children in 37 schools in the city of Groningen, the Netherlands. The authors demonstrate that any simple ranking for the schools would be misleading because of important interactions of student background with school.