ShrinkageTrees

This package provides functions for fitting Horseshoe Trees, Causal Horseshoe Forests, and their more general counterparts: Shrinkage Trees and Causal Shrinkage Forests.

These models allow for global-local shrinkage priors on tree step heights, enabling adaptive modeling in high-dimensional settings.

The functions can be used for:

1. High-dimensional prediction
2. High-dimensional causal inference
3. Estimaton of heterogeneous (conditional average) treatment effects

Supported outcome types: continuous, binary, and right-censored survival times.

The mathematical background and theoretical foundation for these models will appear in "Horseshoe Forests for High-Dimensional Causal Survival Analysis" by T. Jacobs, W.N. van Wieringen, and S.L. van der Pas (2025).

✨ Features

• Horseshoe, forest-wide horseshoe, empirical Bayes Horseshoe, and half-Cauchy priors
• Flexible tree-based non-linear modeling of the ATE and CATE
• Supports survival data with right-censoring (accelerated failure time model)
• Efficient C++ backend via Rcpp