Instrumental Variables in Sparse and Dynamical Settings

on YouTube ↗43m

" summary

Shows that when causal effects are sparse and instruments propagate enough heterogeneity, the true linear beta is the sparsest solution in the solution space of IV moment equations and becomes unique under two concrete conditions: a rank condition on the instruments-to-parents matrix and no exact cancellation of directed paths (graphically: at least as many node-disjoint instrument→parent paths as parents). Gives a counterexample with a bottlenecked three-edge collapse, proposes nuisance IV and conditional-IV fixes for VAR time series, and reports successful applied tests on biological and chemical examples.

AI-written summary · about

" chapters (4)

1. 0:00Start
2. 1:14Instrumental Variables
3. 19:35Third Condition
4. 27:17Conditional Ib

" related talks

• 43mRobustness in High-Dimensional Inference TasksJelena Bradic · Robust and High-Dimensional Statistics (Simons Institute workshop, 2018)
• 53mAdversarial Machine Learning and Instrumental Variables for Flexible Causal ModelingVasilis Syrgkanis · Adversarial Approaches in Machine Learning 2022
• 57mA New Perspective on High-Dimensional Causal InferencePragya Sur · Deep Learning Theory Workshop and Summer School 2022
• 44mInvariance, Causality and Novel RobustnessPeter Bühlmann · Robust and High-Dimensional Statistics (Simons Institute workshop, 2018)
• 33mSimple Yet Efficient Estimators For Network Causal Inference...Christina Yu · Algorithmic Aspects of Causal Inference 2022
• 45mHigh Dimensional Robust Sparse RegressionConstantine Caramanis · Sublinear Algorithms and Nearest-Neighbor Search
• 49mCausal Inference in Complex Systems: Network Interference, Strategic Agents, and BeyondPanos Toulis · Quantifying Uncertainty: Stochastic, Adversarial, and Beyond
• 01h 02mPrediction, Generalization, Complexity: Revisiting the Classical View from Statistics Part 2Ryan Tibshirani · Modern Paradigms in Generalization Boot Camp 2024