Eta
Docs
Counterfactual Inference - std.causal.counterfactual
This module adds pragmatic counterfactual APIs on top of Eta’s ADMG identification engine.
Import
(import std.causal.counterfactual)
API
| Function | Description |
|---|---|
(twin-network g intervened-vars) | Duplicates g into a counterfactual world with primed node names and intervention-aware edge surgery |
(id* g gamma) | Counterfactual identification entry point for conjunctions of Y_X events |
(idc* g gamma delta) | Conditional counterfactual identification entry point |
(do:ett g y x x*) | Effect of treatment on the treated query `P(Y_x |
Event shapes
gamma is a list of counterfactual events:
'((y x)) ; Y_x
'((y (x z))) ; Y_{x,z}
delta is a list of conditioning assignments where the first element is
the variable name:
'((x 0))
'((x 0) (z 1))
Example
(define g '((x -> m) (m -> y) (x <-> y)))
;; Counterfactual identification
(id* g '((y x)))
;; Effect of treatment on the treated
(do:ett g 'y 'x 0)
Notes
id*andidc*normalize counterfactual syntax, then delegate tostd.causal.identify(idandidc) for estimand construction.twin-networkremoves incoming causal mechanisms for intervened variables in the counterfactual copy and keeps cross-world latent links for non-intervened variables.