|
|||||
|
|
Using regression tree ensembles to model interaction effects: a graphical approach |
|
| Authors: | Fritz Schiltz Chiara Masci Tommaso Agasisti Daniel Horn |
| Institution: | 1. Leuven Economics of Education Research, University of Leuven, Leuven, Belgiumfritz.schiltz@kuleuven.be;3. Department of Mathematics, Modelling and Scientific Computing, Politecnico di Milano, Italy;4. Department of Management, Economics and Industrial Engineering, Politecnico di Milano, Italy;5. Centre for Economic and Regional Studies, Hungarian Academy of Sciences, Hungary |
| Abstract: | Multiplicative interaction terms are widely used in economics to identify heterogeneous effects and to tailor policy recommendations. The execution of these models is often flawed due to specification and interpretation errors. This article introduces regression trees and regression tree ensembles to model and visualize interaction effects. Tree-based methods include interactions by construction and in a nonlinear manner. Visualizing nonlinear interaction effects in a way that can be easily read overcomes common interpretation errors. We apply the proposed approach to two different datasets to illustrate its usefulness. |
| Keywords: | heterogeneous effects Regression trees interaction effects machine learning education economics |