Research map / Probabilistic and scientific modeling / Causal discovery
The Factorized Self-Controlled Case Series Method: An Approach for Estimating the Effects of Many Drugs on Many Outcomes
Ramin Moghaddass, Cynthia Rudin, David Madigan
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 |
|---|---|---|---|
| add multiple drugs jointly (MSCCS instead of single-drug SCCS) | SCCS · AIM-94 diabetes, coefficient sign agreement | AUC: 0.693 -> 0.773 | Improved |
| add l2 regularization to MSCCS (ablation) | MSCCS · AIM-94 diabetes, coefficient sign agreement | AUC: 0.773 -> 0.774 | No clear effect |
| add hierarchical Bayesian shrinkage across drugs and outcomes (Model 0) | Regularized MSCCS · AIM-94 diabetes, coefficient sign agreement | AUC: 0.774 -> 0.813 | Improved |
| add one level of latent factors, F=2 (ablation) | Model 0 · AIM-94 diabetes, coefficient sign agreement | AUC: 0.813 -> 0.852 | Improved |
| add one level of latent factors, F=3 (ablation) | Model 0 · AIM-94 diabetes, coefficient sign agreement | AUC: 0.813 -> 0.824 | Improved |
| increase latent factors from 2 to 3 (Model 1) (ablation) | Model 1 · AIM-94 diabetes, coefficient sign agreement | AUC: 0.852 -> 0.824 | Hurt |
| add second latent level (Model 2 vs Model 1, 2 factors) (ablation) | Model 1 (F=2) · AIM-94 diabetes, coefficient sign agreement | AUC: 0.852 -> 0.856 | No clear effect |
| add second latent level (Model 2 vs Model 1, 3 factors) (ablation) | Model 1 (F=3) · AIM-94 diabetes, coefficient sign agreement | AUC: 0.824 -> 0.836 | No clear effect |
| use latent factors instead of full coefficient matrix (ablation) | Model 0 · AIM-94 diabetes, computation | CPU time and MCMC mixing | Improved |
| use latent factors (Model 1 F=2) instead of Model 0 (ablation) | Model 0 · AIM-94 diabetes, outcome: Too low | mean AUC, Too low: 70.30% -> 70.86% | No clear effect |
| use latent factors (Model 1 F=2) instead of Model 0 (ablation) | Model 0 · AIM-94 diabetes, outcome: Hypo Symptom | mean AUC, Hypo Symptom: 59.17% -> 56.71% | Hurt |
Closest papers by meaning
- Time Series Deconfounder: Estimating Treatment Effects over Time in the Presence of Hidden Confounders (ICML 2020 · similarity 0.85)
- A Linear Algebraic Framework for Counterfactual Generation (ICLR 2024 · similarity 0.84)
- Temporal Causal Mediation through a Point Process: Direct and Indirect Effects of Healthcare Interventions (NeurIPS 2023 · similarity 0.84)
- Bayesian Inference of Individualized Treatment Effects using Multi-task Gaussian Processes (NeurIPS 2017 · similarity 0.84)
- PairNet: Training with Observed Pairs to Estimate Individual Treatment Effect (ICML 2024 · similarity 0.83)
- A Unified Framework for the Transportability of Population-Level Causal Measures (NeurIPS 2025 · 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.
