Case Study / 2026

Treatment Effect Estimation

A causal inference pipeline for IHDP treatment-effect estimation, covering naive baselines, S/T/X-learners, manual DML, EconML estimators, policy analysis, and reproducible evaluation artifacts for ATE and CATE.

PythonEconMLDoWhyscikit-learnSHAPPandas

Source code

Project Features

What this case study includes

05 features

Reusable preprocessing pipeline for IHDP with aligned scaling and train/test splits

ATE and CATE estimation with S-Learner, T-Learner, X-Learner, and propensity baselines

Manual DML plus EconML LinearDML and CausalForestDML implementations

Evaluation outputs for PEHE, ATE error, predictions, and summary metrics

Interpretability and policy artifacts including SHAP-based analysis and policy curves

Overview

This project turns notebook-based causal inference work into a reusable Python pipeline for treatment-effect estimation on the IHDP benchmark. The goal is not just prediction accuracy, but estimating what changes under treatment and how that effect varies across units.

What It Covers

Consistent preprocessing, scaling, and aligned train/test splitting
Naive ATE and propensity-score stratification baselines
S-Learner, T-Learner, and X-Learner implementations
Manual DML plus EconML `LinearDML` and `CausalForestDML`
Evaluation outputs for PEHE, ATE error, policy curves, and artifact export

Why It Matters

The project shows applied understanding of causal ML beyond standard predictive modeling. It moves from notebook experimentation into a runnable pipeline with reproducible outputs, evaluation artifacts, and model interpretation support.

Stack

Python
scikit-learn
EconML
DoWhy
SHAP
pandas