Marginal Treatment Effects with Misclassified Treatment
Talkat Quantum Physics and Statistical Causal Models (Simons Institute workshop, 2022) · 2022-04-27..04-29·
summary
Derives identification and tight analytical bounds for the marginal treatment effect when a binary treatment is misclassified, showing MTE = derivative of E[Y | true propensity score] but the propensity score is only partially identified; characterizes the identified set, proves local sign identification under mild regularity, and proves the bounds strictly improve on existing LATE bounds. Results require a continuous instrument and an upper bound on misclassification (bounds blow up near 50% misclassification); includes simulations and an Indonesia schooling application.
AI-written summary · about
chapters (13)
- 0:00Introduction
- 2:34Contribution
- 5:48The model
- 11:54Link between MTE and standard causal effect parameters
- 16:29Examples
- 19:40Assumptions
- 21:04Identification Results
- 26:28Analytical bounds for the MTE
- 34:38Tighter bounds for the LATE
- 35:25Exogenous misclassification: 1V
- 37:23Numerical illustration
- 39:47Empirical illustration: Cameiro et al. (2017)
- 42:58Conclusion
related talks
33mThe Challenge of Valid Evaluations
33mSimple Yet Efficient Estimators For Network Causal Inference...
54mDebiasing Random Forests For Treatment Effect Estimation
41mpMSE Mechanism: Differentially Private Synthetic Data with Maximal Distributional Similarity
30mIndividual Statistical Fairness
27mA Practitioner’S Guide To Intent-To-Treat Effects From Multisite (Blocked)...
27mDreadful Frailties in Propensity Score Matching and How to Fix Them
01h 11mStatistical Limits of Causal Inference