Skip to contents

dr3

Doubly-robust ATE estimators — three routes to one answer.

dr3 is a transparent, teaching-oriented implementation of the average treatment effect (ATE) estimators reconciled in Three Routes to One Answer: Reconciling AIPW, TMLE, and Double Machine Learning for Applied Researchers. AIPW, TMLE, and double machine learning all target the same efficient influence function

D(O) = [A/g(X) - (1-A)/(1-g(X))] (Y - Q_A(X)) + (Q_1(X) - Q_0(X)) - psi

and differ only in how they solve it (average the score vs. targeting) and how the nuisances g(X) = P(A=1|X) and Q_a(X) = E[Y|A=a,X] are fit (full-sample, cross-fit, or double cross-fit). dr3 runs the whole family on one shared Super Learner library, so any difference is estimator-driven, not implementation-driven.

Two choices generate the eight-estimator panel:

full sample single cross-fit double cross-fit
one-step AIPW DML DC-AIPW
targeting TMLE CV-TMLE DC-TMLE

plus the singly-robust baselines (g-computation, IPTW).

Scope

dr3 is for teaching and reproducibility. It is a readable reference engine, cross-checked against the established tmle and AIPW packages — not a replacement for them. It targets the ATE for a binary treatment at a single time point; it is not a general causal-inference framework.

Installation 

# install.packages("remotes")
remotes::install_github("ehsanx/dr3")

Quick start 

library(dr3)

d   <- dr_nhefs()                 # ships in causaldata; nothing to download
fit <- dr_panel(d$Y, d$A, d$X)    # one call -> all eight estimators
fit
plot(fit)

dr_panel() is a thin wrapper: fit the shared nuisances yourself and hand them to any estimator — pairing dr_aipw() with cross-fit nuisances is DML.

nu <- dr_nuisance(d$Y, d$A, d$X, strategy = "crossfit")
dr_aipw(d$Y, d$A, nu)   # DML
dr_tmle(d$Y, d$A, nu)   # CV-TMLE

Datasets for the tutorials

dr3 bundles no data — each dataset is pulled from another package.

Loader Data Source
dr_nhefs() Smoking cessation → weight change (continuous) causaldata
dr_rhc() Right-heart catheterization → 30-day mortality (binary) Hmisc::getHdata("rhc")