CRE

CRE implements the Causal Rule Ensemble method for interpretable discovery and inference of heterogeneous treatment effects via ensemble-of-trees approaches.

Key Features:

• Ensemble-of-trees approach: Extensively explores heterogeneity patterns to identify subgroups where treatments have notably different effects, addressing instability and limited exploration of single-tree methods.
• Multiple IATE estimators: Supports AIPW, Causal Forest, Causal BART, S-Learner, T-Learner, and X-Learner as internal individual average treatment effect estimators.
• Decision rule output: Characterizes heterogeneous treatment effects in terms of interpretable decision rules.
• Flexible algorithm variants: Provides multiple variants of the Causal Rule Ensemble algorithm for different research scenarios.

Scientific Applications:

• Heterogeneous treatment effect discovery: Identifies interpretable subgroups of study populations where treatment effects differ notably from the average treatment effect in health and social sciences.
• Causal inference: Enables data-driven discovery of treatment effect heterogeneity without requiring pre-specified subgroups.

Methodology:

CRE uses recursive mathematical programming for HTE maximization within an ensemble-of-trees framework to generate decision rules characterizing heterogeneous treatment effects.