Docs
std.causal.estimate
Causal effect estimation backends. Input data is a list of observations,
each an alist such as '((x . 1) (z . 0) (y . 2.3)). Treatment X is
binary in {0, 1} unless noted.
(import std.causal.estimate)
ATE estimators
| Symbol | Description |
|---|
(do:ate data y x z) | Default ATE estimator (AIPW). |
(do:ate-gformula data y x z) | G-formula via stratified outcome means. |
(do:ate-ipw data y x z) | IPW with stratified empirical propensity scores. |
(do:ate-aipw data y x z) | Augmented IPW (doubly robust). |
(do:ate-tmle data y x z) | Targeted maximum-likelihood estimator. |
Inference
| Symbol | Description |
|---|
(do:bootstrap-ci estimator data y x z . opts) | Non-parametric bootstrap CI. Options include 'reps, 'alpha, 'seed. |
Sub-routines
| Symbol | Description |
|---|
(do:propensity-score data x z) | Stratified empirical propensity scores. |
(causal:design-matrix data cols) | Build a row-major design matrix. |
(causal:response-vector data y) | Build the outcome vector. |
Sensitivity
| Symbol | Description |
|---|
(do:e-value rr) | Compute the E-value for a risk ratio. |
(do:rosenbaum-bound . args) | Rosenbaum-style bound for an unmeasured confounder. |
Eta