Tree-structured regression for modeling clustered data

Nikolai Spuck, University of Bonn

Co-authors: Matthias Schmid, University of Bonn; Moritz Berger, University of Bonn

Abstract: Tree-structured models are a powerful alternative to parametric regression models if non-linear effects and interactions are present in the data. Yet, classical tree-structured models might not be appropriate if data comes in clusters of units, which requires taking the dependence of observations into account. To address this issue, we present a tree-structured approach for clustered data that achieves a sparse modeling of unit-specific effects and identifies subgroups (based on individual-level covariates) that differ with regard to the outcome. Simulation experiments and an application of the proposed model to analyze quality of life in older adults yielded promising results and illustrated the accessibility of the approach.