Causal Inference: Estimands and Ontologies

This lecture by Mats J. Stensrud from EPFL delves into the fundamental concepts of causal inference, focusing on the importance of committing to an ontology for drawing causal inferences. Stensrud discusses the translation of research questions into formal causal estimands, the conditions for identifying causal estimands, and the estimation of causal effects from data. The lecture explores the significance of statistical and causal ontologies in defining parameters and objects, emphasizing the challenges in selecting appropriate estimands. Stensrud also addresses the implications of deterministic and stochastic counterfactual ontologies in causal inference, highlighting the need for careful consideration in observational and randomized experiments.