py-why/EconML

ALICE (Automated Learning and Intelligence for Causation and Economics) is a Microsoft Research project aimed at applying Artificial Intelligence concepts to economic decision making. One of its goals is to build a toolkit that combines state-of-the-art machine learning techniques with econometrics in order to bring automation to complex causal inference problems. To date, the ALICE Python SDK (econml) implements orthogonal machine learning algorithms such as the double machine learning work of Chernozhukov et al. This toolkit is designed to measure the causal effect of some treatment variable(s) t on an outcome variable y, controlling for a set of features x.

What it solves

EconML provides tools to estimate heterogeneous treatment effects from observational data. It addresses the challenge of determining the causal effect of an intervention (treatment) on an outcome, specifically how that effect varies across different individuals or groups based on their features (heterogeneity).

How it works

The package combines machine learning techniques with econometrics to automate complex causal inference. It allows users to model effect heterogeneity using flexible ML models like random forests, boosting, lasso, and neural networks while maintaining causal interpretation and providing confidence intervals. It supports various estimation methods, including:

  • Double Machine Learning (DML): Uses ML to control for confounders and isolate the treatment effect.
  • Causal Forests and Orthogonal Random Forests: Specialized forest-based methods for estimating effects.
  • Meta-Learners: Implementations of X-Learner, S-Learner, and T-Learner.
  • Doubly Robust Learners: Combines propensity and regression models to improve estimation accuracy.
  • Instrumental Variables (IV): Methods for handling cases where there is an instrument variable to address unobserved confounders.

Who it’s for

It is designed for data scientists and economists who need to perform causal inference on observational datasets to automate personalized decision-making.

Highlights

  • Unified API: Provides a consistent interface for various causal ML methods.
  • Flexible Modeling: Supports integration with standard Python ML packages for modeling effect heterogeneity.
  • Causal Interpretation: Preserves the causal meaning of results and often provides valid confidence intervals.
  • Interpretability Tools: Includes tree-based interpreters and SHAP value support to explain the behavior of the CATE (Conditional Average Treatment Effect) model.

Related

  • Project
    uber/causalml

    A Python package for uplift modeling and causal inference that estimates the causal impact of interventions on individual users to optimize targeting and personalization.

  • Project
    igerber/diff-diff

    A Python library for Difference-in-Differences causal inference that provides a wide range of econometric estimators and diagnostic tools for measuring policy and campaign lift.

  • Project
    microsoft/responsible-ai-toolbox

    A suite of model and data exploration tools that enable developers to identify errors, assess fairness, and interpret AI model predictions for more responsible deployment.

  • Project
    DoubleML/doubleml-for-py

    A Python package implementing the double/debiased machine learning framework for estimating structural parameters with reduced bias using scikit-learn.

  • Project
    microsoft/FLAML

    A lightweight Python library for efficient automated machine learning (AutoML) and hyperparameter tuning to find high-quality models with minimal computational resources.