Reposted by Klaus Pforr
After noticing EconTwitter is basically dead, decided to switch! Welcome any reccomendations on how to use the platform, connect with other colleagues, and best way to promote research over here!
#EconSky #Macroeconomics #LaborEcon
![2026-10-14 | Input talk | Valentyn Melnychuk (LMU Munich)
Causal ML for treatment effect estimation
more
Hybrid event [A5, 6, Room A231 + Zoom]
October 14, 2026, 13:45-15:15
Abstract
In this talk, I introduce causal machine learning as a practical toolkit for estimating treatment effects and counterfactuals from observational data. Here, I formalize the potential-outcomes setup and explain why individual treatment effects are not directly observable, which motivates causal identification. I then walk through the key assumptions — consistency, overlap/positivity, and exchangeability — and show how they connect to common causal frameworks (potential outcomes, SCMs, and causal graphs). Building on that foundation, I cover the main estimands (ATE and heterogeneous effects such as CATE) and compare widely used CATE estimators, from plug-in learners (S-/T-learners) to orthogonal and doubly robust approaches based on pseudo-outcomes and residualization (e.g., IPW/RA/DR and R-learning). I conclude the talk with an overview of state-of-the-art works in causal ML and a discussion of open practical challenges such as uncertainty quantification, model selection under limited data, and hidden confounding.
Presenter(s)
Valentyn Melnychuk is a PhD candidate at the Institute of AI in Management, LMU Munich. His research focuses on causal inference and data-driven decision-making. In particular, he develops methods that combine causal inference with probabilistic modelling, with a particular emphasis on uncertainty quantification, representation learning, generative models, and robustness to confounding or violated assumptions.](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:oizrkjzyr6vekl76tjctghcl/bafkreihj6il7mqbjsspjan3azwfj5a7vysf5iegkvcd77sec4jtw7i6at4)
