An Introduction to Causal Inference

Prof. Michael Sobel

Columbia University

About the Speaker

Michael Sobel is a professor in the statistics department at Columbia University. He has published extensively in the area of social statistics, particularly on structural equation models and categorical data analysis, and is a past editor of Sociological Methodology. His more recent work is in the area of causal inference, where he has published papers on mediation, compliance, interference, and longitudinal data analysis using fixed effects models. His most recent work takes up the subject of making causal inferences for fMRI data.

Abstract

Causal inference as a field in statistics has developed rapidly over the past 30 years. During the seminar, I will compare and contrast several approaches to causation and causal inference including probabilistic causation and Granger causality, and Rubin's causal model. Attention then focuses on Rubin's model, which is based on the idea that causal relationships sustain counterfactual conditional statements. To formalize this, potential outcomes notation is introduced and a simple example is presented. Ignorability conditions under which the observed data can be used to make valid causal inferences are given.