Research map / Probabilistic and scientific modeling / Treatment effects
Naive regression requires weaker assumptions than factor models to adjust for multiple cause confounding
Justin Grimmer, Dean Knox, Brandon Stewart
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 full probabilistic-PCA deconfounder instead of naive regression | naive regression · Medical deconfounder simulation (orig: beta1=0, beta2=0.3) | RMSE beta1: 0.125 -> 0.690 | Hurt |
| use PCA+CV-Ridge deconfounder instead of naive regression when true beta1 is zero | naive regression · Medical deconfounder simulation (orig: beta1=0, beta2=0.3) | RMSE beta1: 0.125 -> 0.031 | Improved |
| use PCA+CV-Ridge deconfounder instead of naive regression when true beta1 is -0.3 | naive regression · Medical deconfounder simulation (ours: beta1=-0.3, beta2=0.3) | RMSE beta1: 0.191 | Hurt |
| switch true effect of A1 from 0 to -0.3 | PCA+CV-Ridge · Medical deconfounder simulation (ours: beta1=-0.3, beta2=0.3) | RMSE beta1: 0.031 -> 0.191 | Hurt |
| add more treatments (m=3 to m=200) for naive regression | naive regression · Subset simulation, beta_j = 10 | avg RMSE: 0.333 -> 0.011 | Improved |
| add more treatments (m=3 to m=200) for subset deconfounder with constant effects | subset deconfounder · Subset simulation, beta_j = 10 | avg RMSE: 10.000 -> 10.000 | No clear effect |
| add more treatments for subset deconfounder with beta_j = 100 | subset deconfounder · Subset simulation, beta_j = 100 | avg RMSE: 100.000 -> 100.000 | No clear effect |
| add more treatments for subset deconfounder with beta_j ~ N(1,2) | subset deconfounder · Subset simulation, beta_j ~ N(1,2) | avg RMSE: 1.465 -> 1.026 | Mixed |
| add more treatments for subset deconfounder with beta_j = 1/m | subset deconfounder · Subset simulation, beta_j = 1/m | avg RMSE: 0.611 -> 0.033 | Improved |
| adjust for budget in actor regression | deconfounder · TMDB 5000 actor effects | plausibility of actor effects | Improved |
| use factor-model substitute confounder instead of plain regression on actors | naive regression · TMDB 5000 actor effects | plausibility of top actor effect: 831% | Hurt |
| use subset deconfounder for single actors | naive regression · TMDB 5000 actor effects | plausibility of top actor effect: 1,128% | Hurt |
Closest papers by meaning
- Sense and Sensitivity Analysis: Simple Post-Hoc Analysis of Bias Due to Unobserved Confounding (NeurIPS 2020 · similarity 0.87)
- Proximal Causal Inference With Text Data (NeurIPS 2024 · similarity 0.86)
- Detecting and Measuring Confounding Using Causal Mechanism Shifts (NeurIPS 2024 · similarity 0.86)
- Data Fusion for Partial Identification of Causal Effects (NeurIPS 2025 · similarity 0.85)
- Conditional Instrumental Variable Regression with Representation Learning for Causal Inference (ICLR 2024 · similarity 0.85)
- Automating the Selection of Proxy Variables of Unmeasured Confounders (ICML 2024 · 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.
