In the paper “Statistical Analysis of Teaching Styles” (Aitkin, Anderson and Hinde, JRSSB, 1981) a collection of modelling techniques was used for the reanalysis of a large educational dataset on teaching style and pupil performance. Specific methods included: a latent class (mixture) model for clustering based on binary indicators of teaching style; variance component models relating the teaching styles, determined by the cluster analysis, to pupil progress in three subject domains; and normal mixture and factor models for pupil personality. There were many common strands to the different models used and in the computational approaches, with heavy reliance on the EM algorithm. As well as providing a substantial reanalysis of an important dataset, the paper highlighted the power of a model-based approach.

All of the models used in this paper have seen many extensions and generalizations. Computational techniques have improved and software is now widely available, in particular, the R system makes it easy to follow and extend the analyses in the paper. We use this to illustrate how such case studies can be used to motivate students in statistical modelling, going beyond standard textbook examples.

Keywords: statistical modelling; R system; teaching case studies

Biography: John Hinde is the Professor of Statistics at the National University of Ireland, Galway having formerly worked at the Universities of Exeter, Lancaster and Newcastle in the UK. His main interests are in statistical modelling, particulary generalized linear models and their random effect extensions for overdispersion. He is a coauthor if Statistical Modelling in R and over 50 publications in applied and methodological statistics.