Causal Inference in the Empirical Sciences

Loading Map....

Date/Time
Date(s) - Feb 3, 2009
3:15 PM - 4:15 PM

Location
Center for Community Health

Category(ies)


Presented by: 

Judea Pearl, Ph.D.

Departments of Computer Science and Statistics University of California, Los Angeles

Abstract:

The questions that motivate most studies in the health, social and behavioral sciences are not statistical but causal in nature. For example, what is the efficacy of a given drug in a given population? Whether data can prove an employer guilty of hiring discrimination? What fraction of past crimes could have been prevented by a given policy? What was the cause of death of a given individual, in a specific incident?

Remarkably, although much of the conceptual and algorithmic tools needed for tackling such problems are now well established, they are hardly known to empirical researchers.

The barrier has been cultural; formulating causal problems mathematically requires certain extensions to the standard mathematical language of statistics, and these extensions are not generally emphasized in the mainstream literature and education. (Skeptics and traditionalists in the audience will be invited to write down a mathematical formula for the empirical claim: ‘The rooster crow does not cause the sun to rise’)

In this talk, I will attempt to break this cultural barrier by introducing a few basic principles and simple mathematical tools that are sufficient for solving most (if not all) problems involving causal relationships. The principles are based on non-parametric structural equation models, a natural generalization of those used by econometricians in the 1950-60s, yet cast in new mathematical underpinnings. This framework, enriched with a few ideas from logic and graph theory, gives rise to a formal yet friendly calculus of counterfactuals that unifies all existing approaches to causation — from econometric and Rubin’s models to path-diagrams — and resolves long-standing problems in several of the sciences. These include questions of confounding, causal effect estimation, covariate selection, policy analysis, legal responsibility, effect decomposition, instrumental variables, and the integration of data from diverse studies.

Reference: J. Pearl, Causality (Cambridge University Press, 2000) http://bayes.cs.ucla.edu/jp_home.html

Background and Tutorials:

http://bayes.cs.ucla.edu/IJCAI99/

ftp://ftp.cs.ucla.edu/pub/stat_ser/Test_pea-final.pdf

ftp://ftp.cs.ucla.edu/pub/stat_ser/R273.pdf

Biography:

Judea Pearl was born in Tel Aviv and is a graduate of the Technion – Israel Institute of Technology. He came to the United States for post graduate work in 1960 and received his Master’s degree in physics from Rutgers University and a PhD degree in 1965 from the Brooklyn Polytechnic Institute. Until 1969, he held research positions at RCA David Sarnoff Research Laboratories in Princeton, New Jersey and Electronic Memories, Inc. Hawthorne, California. Dr. Pearl joined the faculty of UCLA in 1969, where he is currently a Professor of Computer Science and Statistics and Director of the Cognitive Systems Laboratory. He conducts research in artificial intelligence, human reasoning and philosophy of science, and is the author of three landmark books in these areas: Heuristics (1984), Probabilistic Reasoning (1988), and Causality (2000). http://singapore.cs.ucla.edu/BOOK-2K/

A member of the National Academy of Engineering and a Founding Fellow of the American Association for Artificial Intelligence, Professor Pearl is a recipient of numerous scientific awards, including the 2001 London School of Economics Award for the “best book in the philosophy of science”, the 2004 ACM Allen Newell Award for “seminal contributions that extend to philosophy, psychology, medicine, statistics, econometrics, epidemiology and social science.” In April 2008, he received the Benjamin Franklin Medal for Computer and Cognitive Science from the Franklin Institute “for creating the first general algorithms for computing and reasoning with uncertain evidence, allowing computers to uncover associations and causal connections hidden within millions of observations”.