Introduction to Targeted Maximum Likelihood Estimation from scratch

Targeted Maximum Likelihood Estimation (TMLE) is a powerful, flexible framework for drawing causal conclusions from observational data. Unlike traditional statistical methods, TMLE is designed to explicitly target the causal parameter of interest, combining machine learning with rigorous statistical theory to reduce bias and improve efficiency. In this session, we’ll build TMLE from the ground up and explore how it enables us to emulate randomized controlled trials using real-world data, even in the presence of confounding and missingness.