Public Module Reference
CATE Estimators
Double Machine Learning (DML)
|
The base class for parametric Double ML estimators. |
|
The Double ML Estimator with a low-dimensional linear final stage implemented as a statsmodel regression. |
|
A specialized version of the Double ML estimator for the sparse linear case. |
|
A Causal Forest [cfdml1] combined with double machine learning based residualization of the treatment and outcome variables. |
|
The base class for non-parametric Double ML estimators, that can have arbitrary final ML models of the CATE. |
|
A specialized version of the linear Double ML Estimator that uses random fourier features. |
Doubly Robust (DR)
|
CATE estimator that uses doubly-robust correction techniques to account for covariate shift (selection bias) between the treatment arms. |
|
Special case of the |
|
Special case of the |
|
Instance of DRLearner with a |
Meta-Learners
|
Meta-algorithm proposed by Kunzel et al. that performs best in settings |
|
Conditional mean regression estimator. |
|
Conditional mean regression estimator where the treatment assignment is taken as a feature in the ML model. |
Meta-algorithm that uses domain adaptation techniques to account for |
Orthogonal Random Forest (ORF)
|
OrthoForest for continuous or discrete treatments using the DML residual on residual moment function. |
|
OrthoForest for discrete treatments using the doubly robust moment function. |
Instrumental Variable CATE Estimators
Double Machine Learning (DML) IV
|
Implementation of the orthogonal/double ml method for CATE estimation with IV as described in section 4.2: |
|
The base class for parametric DMLIV estimators to estimate a CATE. |
|
The base class for non-parametric DMLIV that allows for an arbitrary square loss based ML method in the final stage of the DMLIV algorithm. |
Doubly Robust (DR) IV
|
The DRIV algorithm for estimating CATE with IVs. |
|
Special case of the |
|
Special case of the |
|
Instance of DRIV with a |
|
Implements the DRIV algorithm for the intent-to-treat A/B test setting |
|
Implements the DRIV algorithm for the Linear Intent-to-Treat A/B test setting |
DeepIV
|
The Deep IV Estimator (see http://proceedings.mlr.press/v70/hartford17a/hartford17a.pdf). |
Sieve Methods
|
Non-parametric instrumental variables estimator. |
|
Featurizer that returns(unscaled) Hermite function evaluations. |
Featurizer that returns the derivatives of |
Estimators for Panel Data
Dynamic Double Machine Learning
|
CATE estimator for dynamic treatment effect estimation. |
Policy Learning
|
Policy learner that uses doubly-robust correction techniques to account for covariate shift (selection bias) between the treatment arms. |
|
Policy learner that uses doubly-robust correction techniques to account for covariate shift (selection bias) between the treatment arms. |
|
Welfare maximization policy forest. |
|
Welfare maximization policy tree. |
CATE Interpreters
An interpreter for the effect estimated by a CATE estimator |
|
An interpreter for a policy estimated based on a CATE estimation |
CATE Validation
|
Validation tests for CATE models. |
|
Results class for BLP test. |
Results class for calibration test. |
|
Results class for uplift curve-based tests. |
|
|
Results class for combination of all tests. |
CATE Scorers
|
Scorer based on the RLearner loss. Fits residual models at fit time and calculates residuals of the evaluation data in a cross-fitting manner::. |
A CATE estimator that represents a weighted ensemble of many CATE estimators. |
Generalized Random Forests
|
A Causal Forest [cf1]. |
|
A Causal IV Forest [cfiv1]. |
|
An implementation of a subsampled honest random forest regressor on top of an sklearn regression tree. |
|
Simple wrapper estimator that enables multiple outcome labels for all the grf estimators that only accept a single outcome. |
A criterion class that estimates local parameters defined via linear moment equations of the form. |
|
|
Base class for Genearlized Random Forests for solving linear moment equations of the form. |
|
A tree of a Generalized Random Forest [grftree1]. This method should be used primarily through the BaseGRF forest class and its derivatives and not as a standalone estimator. It fits a tree that solves the local moment equation problem::. |
Scikit-Learn Extensions
Linear Model Extensions
Debiased Lasso model. |
|
|
Debiased MultiOutputLasso model. |
|
Estimator of a linear model where regularization is applied to only a subset of the coefficients. |
|
Class which mimics weighted linear regression from the statsmodels package. |
|
Class which mimics robust linear regression from the statsmodels package. |
Version of sklearn Lasso that accepts weights. |
|
|
Version of sklearn LassoCV that accepts weights. |
|
Version of sklearn MultiTaskLassoCV that accepts weights. |
|
Helper class to wrap either WeightedLassoCV or WeightedMultiTaskLassoCV depending on the shape of the target. |
Model Selection Extensions
|
An extension of GridSearchCV that allows for passing a list of estimators each with their own parameter grid and returns the best among all estimators in the list and hyperparameter in their corresponding grid. |
|
K-Folds cross-validator for weighted data. |
|
Stratified K-Folds cross-validator for weighted data. |
Inference
Inference Results
|
Results class for inference assuming a normal distribution. |
Results class for inference with an empirical set of samples. |
|
Population summary results class for inferences. |
Inference Methods
Inference instance to perform bootstrapping. |
|
Inference based on predict_interval of the model_final model. |
|
|
Inference based on predict_interval of the model_final model. |
Inference based on predict_interval of the model_final model. |
|
|
Stores statsmodels covariance options. |
Assumes estimator is fitted on categorical treatment and a separate generic model_final is used to fit the CATE associated with each treatment. |
|
Inference method for estimators with categorical treatments, where a linear in X model is used for the CATE associated with each treatment. |
|
Special case where final model is a StatsModelsLinearRegression |
Federated Estimation
A class for federated learning using LinearDML, LinearDRIV, and LinearDRLearner estimators. |
Solutions
Causal Analysis
Note: this class is experimental and the API may evolve over our next few releases. |
Integration with DoWhy
|
A wrapper class to allow user call other methods from dowhy package through EconML. |
Utilities
Utility methods. |
Private Module Reference
Orthogonal Machine Learning is a general approach to estimating causal models by formulating them as minimizers of some loss function that depends on auxiliary regression models that also need to be estimated from data. |
|
Base classes for all CATE estimators. |
|
The R Learner is an approach for estimating flexible non-parametric models of conditional average treatment effects in the setting with no unobserved confounders. |
|
Bootstrap sampling. |