Stanford University
POLISCI 353: Workshop in Statistical Modeling
Jonathan Wand, Simon Jackman, and Doug Rivers

Winter 2004 Calendar

Overview

During the Winter quarter we will be focusing on methods for drawing causal inference from experimental and observational studies. For students writing research papers, there will also be two teaching sessions devoted to developing research projects.

Meeting times and location

The workshop will meet weekly each Monday 3:15-5:05 in Room 400 (the Graham Stuart Lounge) on the 4th floor of Encina Hall West, except where noted below.

Description of weeks

Speaker: Jonathan Wand

Topic: Introduction to causal inference

Tuesday 6:15-8:05 Room 400 Encina Hall West ( NOTE: Tuesday session, Monday is MLK Day)

Speaker: Simon Jackman

Topic: Can Technology Empower Voters?: Applying a Matching Estimator to an Experiment with Incomplete Compliance

Additional reading: Semiparametric Estimation of Average Treatment Effects Under Exogeneity: A Review by Guido Imbens (local copy)

Speaker: Jonathan Wand

Topic: Alternatives to unconfoundedness

Speaker: David Freedman, Department of Statistics, UC Berkeley

Topic: Reflections on Causal Models (handout material)

Related article: Structural equation models: A critical review and his articles on the reading list.

Book length: Statistical Models: Theory and Practice

Speaker: Ed Vytlacil

Topic: Estimating the Returns to Education when it Varies Across Individuals

Methods are drawn from the related paper, Understanding What Instrumental Variables Estimate - Estimating Marginal and Average Returns to Education

Professor Vytlacil also has a number of papers of particular relevance to the workshop available on his webpage:

http://www.stanford.edu/~vytlacil/papers.html

In particular

"Structural Equations, Treatment Effects, and Econometric Policy Evaluation"

"Independence, Monotonicity, and Latent Index Models: An Equivalence Result"

Speaker: Donald Green, Department of Political Science, Yale University

Topic: The Illusion of Learning from Observational Data (ms-word)

If you are unable to read the original ms-word document, here is a pdf distilled version

Related reading: Reclaiming the Experimental Tradition in Political Science

Speaker: David Lee, Department of Economics, UC Berkeley

Topic: Regression Discontinuity Analyses of Elections: Randomized Experiments from Non-random selection

Professor Lee talking about these methods in several different contexts, drawing on the following three papers,

Randomized Experiments from Non-random Selection in U.S. House Elections

Do Voters Affect or Elect Policies? Evidence from the U.S. House

Do Unions Cause Business Failures?

Students will have the opportunity to present empirical research. Ideally for each paper there will be at least one proponent who will present evidence that the study is useful for drawing causal, and at least one critic who will seek to present evidence to the contrary.

Return to POLISCI 353 Homepage

Revised January 8, 2004