cleverly.LearnedRuleValue

class cleverly.LearnedRuleValue(rule: ~cleverly.interventions.learned.LearnedRule = <factory>)

Request the fold-average value of a treatment rule learned inside each training fold.

The target is data-adaptive. It is the average over the outer folds of the value of the rule learned on each fold’s training rows, so it depends on the realized split and on the fitted rules. It is not the value of one rule fitted on all rows, and it is not the value of the optimal rule. Van der Laan and Luedtke (2015), Section 7 and Appendix B, define it and its cross-validated TMLE. The interval can under-cover when the effect is zero or near zero for a share of units, where no limiting rule exists.

The fit needs the fold-evaluated CV-TMLE, CrossFitting(enabled=True, fold_evaluation=True). The default CrossFitting() refuses before any learner and names that remedy.

Parameters:

ruleLearnedRule

The rule class, and the label the reported parameter carries.

See also

cleverly.interventions.LearnedRule

The rule this estimand holds.

RegimeMean

The value of a known rule, fixed before the fit.

cleverly.CrossFitting

The fold-evaluated configuration this estimand needs.

Examples

>>> from cleverly import LearnedRuleValue
>>> estimand = LearnedRuleValue()
>>> estimand.name, estimand.rule.name
('ey_learned_rule', 'learned rule')

Methods

Attributes

LearnedRuleValue.definition: str = "average over the outer folds of the value of the rule learned on each fold's training rows"

LearnedRuleValue.name: str = 'ey_learned_rule'

LearnedRuleValue.rule: LearnedRule