Causal inference in the social sciences: Estimating the effects of time-varying treatments in the presence of time-varying confounding: An application to neighborhood effects on high school graduation

Copy the text below to embed this resource

Date
2013-03-22 (Creation date: 2013-03-22)
Main contributor
David J. Harding
Summary
Conventional regression methods fail when estimating the effects of time-varying treatments in the presence of time-varying confounding. I will discuss methods for causal inference in such situations, focusing on inverse probability of treatment weighting for estimating marginal structural models (Robins et al. 2000) and a two-stage regression with residuals method (Almirall et al. 2010) for estimating time-varying effect moderation in a structural nested mean model. I illustrate the use of these methods in two recent studies of neighborhood effects on high school graduation.
Publisher
IU Workshop in Methods
Collection
Workshop in Methods
Unit
Social Science Research Commons
Notes

Performers

Dr. Harding is Associate Professor of Sociology and Public Policy at the University of Michigan. He studies urban poverty and inequality, incarceration and prisoner reentry, adolescence, and statistical methods for causal inference. His book, Living the Drama (University of Chicago Press, 2010), examines the role of neighborhoods in adolescent outcomes related to education and romantic and sexual behavior, focusing on exposure to violence and the cultural context of poor communities. Harding is currently working on projects on prisoner reentry, the effects of community context on adolescent and young adult romantic relationships, and for-profit colleges and educational inequality.

Access Restrictions

This item is accessible by: the public.