AI Paper Atlas

Research map / Probabilistic and scientific modeling / Treatment effects

Valid Causal Inference with (Some) Invalid Instruments

Jason S Hartford, Victor Veitch, Dhanya Sridhar, Kevin Leyton-Brown

ICML 2021 · Probabilistic and scientific modeling › Treatment effects · approach: Instrumental variable estimation · 6 citations (OpenAlex)

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 changedWhereMeasuredEffect
Aggregate per-instrument DeepIV estimates by the modal intervalDeepIV ensemble · Biased demand simulation with invalid instrumentsmean squared errorImproved
Use the mean of per-instrument DeepIV estimates instead of the modal intervalDeepIV ensemble · Mendelian randomization simulationaverage absolute CATE biasHurt
Treat all candidate instruments as valid rather than filtering by the modal intervalDeepIV · Mendelian randomization simulation with invalid instrumentsaverage absolute CATE biasHurt
Use the modal interval with V set to the true number of valid instruments (ablation)ModeIV · Biased demand simulationmean squared errorImproved
Set V too small (ablation)ModeIV with only 5 valid instruments · Biased demand simulationmean squared errorHurt
Use more conservative choices of V (ablation)ModeIV · Biased demand simulationmean squared errorNo clear effect
Use ModeIV instead of oracle DeepIV on known valid instrumentsDeepIV · Mendelian randomization simulationaverage absolute CATE bias: 0.035 vs 0.037 (50% valid); 0.024 vs 0.032 (100% valid)Hurt

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy