Causal inference

Structure-agnostic Optimality of Doubly Robust Learning for Treatment Effect Estimation

We show that first-order debiasing of black-box ML estimators is optimal for estimating average treatment effect.

Jikai Jin, Vasilis Syrgkanis