Overview
Carolin Strobl is a statistician (Professor of Psychological Methods, University of Zurich) known for work on recursive partitioning, random forests, and model-based trees, and for exposing and correcting selection bias in tree-based variable importance.
Key Contributions / Features
- Bias in random-forest variable importance (Strobl et al. 2007): lead author showing that Gini/bootstrap importance favours high-cardinality and finely-measured predictors, with a conditional-inference-forest + subsampling-without-replacement remedy.
- Contributor to conditional-inference trees and forests (
party/partykit ecosystem) and to model-based recursive partitioning.
Related