Docs

std.causal.cate

Meta-learners for conditional average treatment effects (S/T/X/R/DR).

A learner spec is an alist:

'((fit . (lambda (X y) ... model))
  (predict . (lambda (model X) ... y-hat))
  (kind . regressor)              ; or 'classifier
  (fit-weighted . (lambda (X y w) ... model)))   ; optional

Rows are alists; matrices are row-major lists.

(import std.causal.cate)

Fit

Symbol	Description
`(cate:fit-s-learner data y x cols spec)`	S-learner.
`(cate:fit-t-learner data y x cols spec)`	T-learner.
`(cate:fit-x-learner data y x cols spec)`	X-learner.
`(cate:fit-r-learner data y x cols spec)`	R-learner.
`(cate:fit-dr-learner data y x cols spec)`	DR-learner.

Use

Symbol	Description
`(cate:predict model rows cols)`	Predict CATE for new rows.
`(cate:ate model rows cols)`	Average of predicted CATE.
`(cate:rank model rows cols)`	Rank rows by predicted CATE descending.

Diagnostics

Symbol	Description
`(cate:residual-r2 model data y x cols)`	Residualised R-squared.
`(cate:propensity-overlap data x cols)`	Propensity-score overlap diagnostic.