Part III · Advanced Causal Machine Learning

Module 07. Doubly Robust Estimation & Double Machine Learning

Combine flexible machine learning with orthogonal scores and cross-fitting to get valid causal inference.

Public health: high-dimensional controls (NHEFS smoking cessation) about 6 hours3 lessons

Module progress0/3 lessons · 0%

Lessons

Doubly Robust Estimation: AIPW

· Estimate an ATE by regression adjustment (the outcome-model approach)
· Combine the outcome model and the propensity score into the augmented IPW (AIPW) estimator
· Show by simulation that AIPW is consistent if either model is right ("double robustness")

· Rebuild the partially linear model as "FWL with machine learning"
· Show that using the same data to fit and to residualise creates bias, and fix it with cross-fitting
· Implement a cross-fitted DML estimator by hand with random forests

· Estimate partially linear (PLR) and interactive (IRM) models with the DoubleML library
· Choose and compare nuisance learners (Lasso, random forests, LightGBM) by out-of-sample loss
· Estimate the effect of quitting smoking on weight gain with the NHEFS data