ABSTRACT
A variety of new statistical methods from the field of machine learning have the potential to offer new stimuli for research in the social, educational, and behavioral sciences. In this chapter we focus on one of these methods: model-based recursive partitioning. This algorithmic approach is reviewed and illustrated by means of instructive examples and an application to the Mincer equation, which is commonly used to describe the association between education, job experience, and income in econometric and sociological research. For readers unfamiliar with algorithmic methods, the explanation starts with the introduction of the predecessor method classification and regression trees. As opposed to classification and regression trees that search for groups of observations that differ in the values of a response variable, model-based recursive partitioning searches for groups differing in their estimated parameters of a postulated statistical model. With respect to the application and interpretation of model-based recursive partitioning, we highlight the principle of parsimony and Ockham’s Razor. To facilitate applicability in the social sciences, we close with a section on model-based recursive partitioning software available in the free R system for statistical computing. In addition, a supplement with worked examples for classification and regression trees as well as model-based recursive partitioning is provided on the book’s website as a hands-on tutorial.
