How reubebe/econml works
EconML is a Python toolkit for estimating heterogeneous treatment effects from observational data. The principal workflow is to provide outcomes, treatments and covariates to a CATE estimator, fit its causal model, then inspect effects and—where supported—uncertainty. Estimator families cover orthogonal/DML, doubly robust, instrumental-variable, forest and meta-learner approaches; separate capabilities support interpretation, validation, policy learning and model selection. Shared orthogonal-learning, estimator, inference and sklearn-extension layers underpin this surface. Some feature branches are documented or indexed but their implementation was not sampled, so their internal links are omitted.
Main entry points: _ortho_learner.py Orthogonal learner · _cate_estimator.py CATE interface · dml.py DML estimators · _drlearner.py DR estimators
Components 21
Causal estimation
- DML estimators
econml/dml/dml.py - DR estimators
econml/dr/_drlearner.py - IV estimators
econml/iv - Forest estimators
econml/grf/classes.py - Meta-learners
econml/metalearners/_metalearners.py
Shared modeling
- Orthogonal learner
econml/_ortho_learner.py - CATE interface
econml/_cate_estimator.py - Linear models
econml/sklearn_extensions/linear_model.py - Model selection
econml/sklearn_extensions/model_selection.py - Model utilities
econml/utilities.py
Inference and interpretation
- Inference results
econml/inference/_inference.py - Bootstrap inference
econml/inference/_bootstrap.py - SHAP explanations
econml/_shap.py - CATE interpreter
econml/cate_interpreter/_interpreters.py - Effect validation
econml/validate/drtester.py - Sensitivity analysis
econml/validate/sensitivity_analysis.py
Validation and decisions
- Policy learning
econml/policy/_drlearner.py - Causal analysis
econml/solutions/causal_analysis/_causal_analysis.py - Causal scoring
econml/score/rscorer.py
Other components
- Analyst
- Observed data
Connections 27
- Analyst → Observed data: provides
README.md - Observed data → DML estimators: fits from
README.md - Observed data → DR estimators: fits from
README.md - Observed data → IV estimators: fits from
README.md - Observed data → Forest estimators: fits from
README.md - Observed data → Meta-learners: fits from
README.md - DML estimators → Orthogonal learner: uses
econml/dml/dml.py - DR estimators → Orthogonal learner: uses
econml/dr/_drlearner.py - IV estimators → Orthogonal learner: uses
econml/iv/dr/_dr.py - Forest estimators → Model utilities: uses
econml/grf/classes.py - DML estimators → Linear models: uses
econml/dml/dml.py - DR estimators → Linear models: uses
econml/dr/_drlearner.py - Orthogonal learner → CATE interface: extends
econml/_ortho_learner.py - Orthogonal learner → Model selection: uses
econml/_ortho_learner.py - DML estimators → CATE interface: implements
econml/dml/dml.py - DR estimators → CATE interface: implements
econml/dr/_drlearner.py - CATE interface → Inference results: produces
econml/_cate_estimator.py - Analyst → Inference results: requests intervals
README.md - Analyst → SHAP explanations: requests explanations
README.md - Analyst → CATE interpreter: interprets effects
README.md - Analyst → Effect validation: validates effects
README.md - Analyst → Sensitivity analysis: checks sensitivity
- Analyst → Policy learning: learns policies
README.md - Analyst → Causal analysis: runs analysis
econml/solutions/causal_analysis/_causal_analysis.py - Analyst → Causal scoring: selects models
README.md - CATE interface → SHAP explanations: explains effects
econml/_cate_estimator.py - Effect validation → Sensitivity analysis: shares validation