Expand description
Best linear predictor (BLP) of a CATE and group average treatment effects (GATE), following Semenova & Chernozhukov (2021) and DoubleML.
Given an orthogonal signal φᵢ with E[φ | X] = τ(X) (e.g. the AIPW
pseudo-outcome), the BLP is the OLS fit of φ on a basis b(X), with
heteroskedasticity-robust (HC0) covariance Ω. Pointwise inference for
τ̂(x) = b(x)ᵀβ̂ uses se(x) = √(b(x)ᵀ Ω b(x)).
Structs§
- Blp
- A fitted best linear predictor, one OLS fit per repetition of the signal.