Research map / Probabilistic and scientific modeling / Treatment effects
Debiasing Machine Learning Predictions for Causal Inference Without Additional Ground Truth Data: “One Map, Many Trials” in Satellite-Driven Poverty Analysis
Markus B. Pettersson, Connor Thomas Jerzak, Adel Daoud
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 |
|---|---|---|---|
| apply Tweedie's correction instead of naive predictions | ML wealth predictions · Simulated treatment-effect estimation | MAE: 0.48 -> 0.04 | Improved |
| apply Tweedie's correction instead of naive predictions | ML wealth predictions · Simulated treatment-effect estimation | calibration slope: 0.535 -> 0.995 | Improved |
| apply LCC instead of naive predictions | ML wealth predictions · Simulated treatment-effect estimation | calibration slope: 0.535 -> 1.008 | Improved |
| train with Ratledge loss instead of naive MSE | ResNet-50 predictions · Simulated treatment-effect estimation | calibration slope: 0.535 -> 0.641 | Mixed |
| apply Tweedie's correction instead of naive | ML wealth predictions · DHS ADM2 region means | MAE: 2.67 -> 2.39 | Improved |
| restrict to regions with at least 100 clusters (ablation) | Tweedie's correction · DHS ADM2 region means | calibration slope: 0.83 to 0.90 -> 0.99 | Improved |
| apply Tweedie's correction instead of naive | ML wealth predictions · Aid intervention ATE (DHS) | correlation with ground-truth ATE: 0.958 -> 0.998 | Improved |
| apply Tweedie's correction instead of naive | ML wealth predictions · Aid intervention ATE (DHS) | MAE: 0.69 -> 0.57 | Improved |
| add PPI with 10% fresh downstream data | naive predictions · Aid intervention ATE (DHS) | correlation with ground-truth ATE: 0.958 -> 0.981 | Improved |
| increase fresh downstream data for PPI (ablation) | PPI · Aid intervention ATE (DHS) | MAE | Improved |
| shuffle more training/calibration labels (0 -> 0.30) (ablation) | naive predictions · Aid ATE (label shuffling) | MAE: 0.69 -> 3.91 | Hurt |
| shuffle more training/calibration labels (0 -> 0.30) (ablation) | Tweedie's correction · Aid ATE (label shuffling) | MAE: 0.57 -> 1.32 | Hurt |
Closest papers by meaning
- IncomeSCM: From tabular data set to time-series simulator and causal estimation benchmark (NeurIPS 2024 · similarity 0.87)
- SpaCE: The Spatial Confounding Environment (ICLR 2024 · similarity 0.85)
- An Analysis of Causal Effect Estimation using Outcome Invariant Data Augmentation (NeurIPS 2025 · similarity 0.85)
- Comparing Targeting Strategies for Maximizing Social Welfare with Limited Resources (ICLR 2025 · similarity 0.85)
- Marginal Causal Flows for Validation and Inference (NeurIPS 2024 · similarity 0.84)
- Data Fusion for Partial Identification of Causal Effects (NeurIPS 2025 · similarity 0.84)
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.
