Causal Inference via Machine Learning

Workshop

Slides

Date: February 13, 2026 10:00 AM – 13:00 PM
Event: Data Lab for Social Good Training Session
Location: Aberconway Building Meeting Room, Cardiff University, UK

Who is the course for?

• Researchers and practitioners interested in understanding causal relationships in observational data
• Analysts working with high-dimensional datasets where traditional regression methods may fail
• Anyone seeking to combine modern machine learning techniques with rigorous causal inference frameworks
• Policy analysts and evaluators who need to estimate treatment effects from non-experimental data
• Those familiar with basic statistics who want to advance to cutting-edge methods in econometrics and causal ML

What will you learn?

• The fundamental distinction between correlation and causation, and why it matters for real-world decision-making
• Core identification strategies: RCTs, IV, DiD, and their assumptions.
• The Double Lasso method for valid causal inference when the number of control variables is large
• Double/Debiased Machine Learning (DML) for estimating causal effects in partially linear and fully interactive models
• How to estimate heterogeneous treatment effects and local average treatment effects (LATE) using modern ML methods
• Practical implementation skills through hands-on examples with real datasets

Course Outline

• Motivation and Conceptual Foundations
• Identification
• Randomised Controlled Trials (RCTs)
• Introduction to Linear Regression
• Regularization Techniques for Linear Regression in High-Dimensional Settings
• Statistical Inference on Causal Effects in High Dimensional Linear Regression Models
• Statistical Inference on Predictive and Causal Effects in Modern Nonlinear Regression Models

Prerequisites

• Basic understanding of statistics and probability theory
• Familiarity with linear regression and ordinary least squares (OLS)
• Working knowledge of statistical concepts such as expected values, variance, and correlation
• Basic programming skills (R)
• Understanding of matrix notation and basic linear algebra
• No prior knowledge of causal inference required – the course builds from foundational concepts Helpful but not required: exposure to econometrics, machine learning basics, or policy evaluation methods