Torsten Hothorn

Overview

Torsten Hothorn is a biostatistician (Professor, University of Zurich) and a central figure in the R ecosystem for recursive partitioning and distributional regression — principal author of conditional-inference trees/forests (party/partykit) and of transformation-model and multiple-testing packages.

Key Contributions / Features

• Conditional-inference trees and forests (ctree/cforest): tree induction that separates variable selection from split-point choice via permutation significance tests, giving unbiased variable selection — the basis of the Strobl et al. (2007) remedy for biased random-forest variable importance.
• Coauthor on the random-forest variable-importance bias paper; broad contributions to model-based recursive partitioning, transformation models, and simultaneous inference in R.

Related

• Classification and Regression Trees (CART)
• Strobl et al. (2007) Bias in Random Forest Variable Importance Measures
• Carolin Strobl