Research map / Learning theory and algorithms / Prediction and decision losses
Tree-Values: Selective Inference for Regression Trees
Anna C. Neufeld, Lucy L. Gao, Daniela M. Witten
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use selective Z-intervals instead of naive Z-intervals | CART tree inference · Simulation, level 2 sibling difference | coverage of 95% CI: 0.410 -> 0.951 | Improved |
| use selective Z-intervals instead of naive Z-intervals | CART tree inference · Simulation, level 2 region mean | coverage of 95% CI: 0.645 -> 0.950 | Improved |
| use selective Z-intervals instead of sample splitting | CART tree inference · Simulation, level 1 region mean | coverage of 95% CI: 0.918 -> 0.951 | Improved |
| use selective Z-test instead of naive Z-test for p-values | CART tree inference · Global null a=b=0 | p-value uniformity | Improved |
| estimate sigma with conservative estimator instead of known sigma (ablation) | Selective Z-intervals · Simulation, level 1 region mean | coverage of 95% CI: 0.95 -> 0.98 | Hurt |
| estimate sigma with SSE estimator instead of known sigma (ablation) | Selective Z-intervals · Simulation, level 2 region mean | coverage of 95% CI: 0.95 -> 0.94 | Hurt |
| use conservative sigma estimator for selective Z-test (ablation) | Selective Z-test · Simulation power, large b | power | Hurt |
| use SSE sigma estimator instead of known sigma (ablation) | Selective Z-test · Simulation power | power | No clear effect |
| condition on ancestors of the region (extra information) (ablation) | Selective Z-test for regions · Simulation (appendix) | power | No clear effect |
| apply selective inference to CART instead of CTree p-value | Box Lunch Study first split hunger<1.8 · Box Lunch Study | p-value: less than 0.001 -> 0.44 | Mixed |
Closest papers by meaning
- Tree-Based Models for Correlated Data (JMLR 2022 · similarity 0.85)
- Partition Tree: Conditional Density Estimation over General Outcome Spaces (NeurIPS 2026 · similarity 0.83)
- Post-ADC Inference: Valid Inference After Active Data Collection (NeurIPS 2026 · similarity 0.83)
- Prediction-Augmented Trees for Reliable Statistical Inference (NeurIPS 2026 · similarity 0.83)
- Piecewise Constant and Linear Regression Trees: An Optimal Dynamic Programming Approach (ICML 2024 · similarity 0.82)
- Multistage Defer Trees for Hybrid Interpretability: If at First You Can't Succeed, Tree Again (NeurIPS 2026 · similarity 0.82)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
