Research map / Probabilistic and scientific modeling / Treatment effects
Validating Causal Inference Models via Influence Functions
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Select model per dataset with IF-based validation instead of single best fixed model | IF-based validation · 77 ACIC datasets | PEHE: 47% smaller | Helped |
| Use IF-based validation instead of plug-in validation | IF-based validation · 77 ACIC datasets | % winner: 65 % -> 72 % | Helped |
| Use IF-based validation instead of IPTW validation | IF-based validation · 77 ACIC datasets | % winner: 54 % -> 72 % | Helped |
| Use IF-based validation instead of factual validation | IF-based validation · 77 ACIC datasets | % winner: 53 % -> 72 % | Helped |
| Use IF-based validation instead of random selection | IF-based validation · 77 ACIC datasets | % winner: 10 % -> 72 % | Helped |
| Compare IF-based validation with supervised cross-validation seeing counterfactuals | IF-based validation · 77 ACIC datasets | % winner: 72 % -> 84 % | Hurt |
| Add influence-function correction to plug-in estimate (ablation) | plug-in PEHE estimate · 77 ACIC datasets | model selection frequency | Helped |
| Add influence-function correction to plug-in estimate (ablation) | plug-in PEHE estimate · 77 ACIC datasets | PEHE estimate bias and variance | Helped |
| Vary number of trees in XGBoost plug-in model (ablation) | IF-based validation · 77 ACIC datasets | probability of selecting best model | No clear effect |
| Use higher-order influence terms (m>1) instead of m=1 (ablation) | IF-based validation · 77 ACIC datasets | PEHE | No clear effect |
Closest papers by meaning
- Causal Estimation with Functional Confounders (NeurIPS 2020 · similarity 0.89)
- Marginal Causal Flows for Validation and Inference (NeurIPS 2024 · similarity 0.88)
- Generalized Bayes for Causal Inference (NeurIPS 2026 · similarity 0.88)
- Unified Covariate Adjustment for Causal Inference (NeurIPS 2024 · similarity 0.87)
- Proximal Causal Inference With Text Data (NeurIPS 2024 · similarity 0.87)
- Doubly Robust Proximal Causal Learning for Continuous Treatments (ICLR 2024 · similarity 0.87)
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.