Research map / Probabilistic and scientific modeling / Treatment effects
Overlap-weighted orthogonal meta-learner for treatment effect estimation over time
Konstantin Hess, Dennis Frauen, Mihaela van der Schaar, Stefan Feuerriegel
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 WO-learner instead of best baseline at lowest overlap | transformer meta-learners · D_gamma, gamma=0.5 | CATE RMSE: 54.4% | Improved |
| use WO-learner instead of best baseline at gamma=1.0 | transformer meta-learners · D_gamma, gamma=1.0 | CATE RMSE: 58.4% | Improved |
| use WO-learner instead of best baseline at highest gamma (lowest overlap) | transformer meta-learners · D_gamma, gamma=6.5 | CATE RMSE: 13.5% | Improved |
| use WO-learner instead of best baseline with n_train=5000 | transformer meta-learners · D_N low-sample | CATE RMSE: 66.9% | Improved |
| use WO-learner instead of best baseline with n_train=2000 | transformer meta-learners · D_N low-sample | CATE RMSE: 28.7% | Improved |
| use WO-learner instead of best baseline at horizon 4 | transformer meta-learners · MIMIC-III semi-synthetic | CATE RMSE: 28.5% | Improved |
| use WO-learner instead of best baseline at horizon 2 | transformer meta-learners · MIMIC-III semi-synthetic | CATE RMSE: 23.1% | Improved |
| swap transformer backbone for LSTM (ablation) | WO-learner · D_gamma, gamma=0.5 | CATE RMSE: 0.03 -> 0.07 | Hurt |
| use WO-learner with LSTM backbone at gamma=1.0 (ablation) | WO-learner · D_gamma, gamma=1.0 | CATE RMSE: -7.8% | No clear effect |
| use WO-learner with LSTM backbone at gamma=1.5 to 6.5 (ablation) | LSTM meta-learners · D_gamma, gamma>=1.5 | CATE RMSE: 46.0% at gamma 1.5; 74.7% at gamma 4.0 | Improved |
| add overlap weights and Neyman-orthogonality (WO) versus IPW at gamma=4.0 | transformer · D_gamma, gamma=4.0 | CATE RMSE: 0.70 -> 0.10 | Improved |
| add Neyman-orthogonal WO relative to RA-learner as covariate dimension grows | transformer · D_mu complex response | CATE RMSE | Improved |
Closest papers by meaning
- Model-agnostic meta-learners for estimating heterogeneous treatment effects over time (ICLR 2025 · similarity 0.91)
- Orthogonal Survival Learners for Estimating Heterogeneous Treatment Effects from Time-to-Event Data (NeurIPS 2025 · similarity 0.91)
- A Meta-learner for Heterogeneous Effects in Difference-in-Differences (ICML 2025 · similarity 0.87)
- Meta-Learners for Partially-Identified Treatment Effects Across Multiple Environments (ICML 2024 · similarity 0.87)
- Quantifying Aleatoric Uncertainty of the Treatment Effect: A Novel Orthogonal Learner (NeurIPS 2024 · similarity 0.86)
- ODE Discovery for Longitudinal Heterogeneous Treatment Effects Inference (ICLR 2024 · similarity 0.86)
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.
