Causal ML

ML-powered causal estimation with honest inference (Athey–Imbens–Wager + Chernozhukov schools): heterogeneous treatment effects via honest causal forests/GRF and the R-learner with RATE/Qini evaluation; double/debiased ML (Neyman orthogonality + cross-fitting) with omitted-variable sensitivity bounds; policy learning from AIPW scores under budget constraints with off-policy evaluation and adaptive-experiment corrections; and panel methods (synthetic control, SDID, matrix completion, staggered-DiD traps). Use whenever a treatment/program/policy EFFECT is estimated with ML anywhere in the pipeline — "who benefits from training", "did the program work, observationally", "ass…