Estimating Causal Effects of Continuous Exposures

Donna Coffman - Research Associate Professor and Principal Investigator, The Methodology Center, Penn State University

Tuesday, October 13, 2015
Time: 11:00 AM – 12:00 PM Pacific / 2:00 PM – 3:00 PM Eastern
Host Location: Santa Monica, conference room 3312
Other Locations: Pittsburgh (room 6202), Washington, DC (room 7401), & Boston (room 9005)

Abstract

In this talk I will provide background to define causal effects for continuous treatments or exposure variables. I will then discuss propensity score estimation for continuous exposures, specifically the use and performance of generalized boosted modeling (GBM) in comparison to linear regression. I will present two simulation studies. The first will evaluate the assessment of balance in the context of a continuous exposure. The second will evaluate approaches for handling missingness in the covariates when estimating propensity scores for continuous exposures. The proposed methodology is illustrated using the Early Dieting in Girls Study, which is a longitudinal study that examined parental influences on daughters' growth and development from ages 5 - 15. In this analysis, we examine the effect of mothers' overall weight concerns and eating behaviors on girls' early dieting behavior.

About the Presenter

Donna Coffman is interested in behavioral health risk factors, namely, diet, substance use, stress, and physical activity (both exercise and reducing sedentary behavior), the causal antecedents of engagement in these behaviors, and the causal effects of these behaviors on health outcomes. Her goal is to examine dynamic causal processes of health behavior change and maintenance. Ultimately, she hopes to inform the development of individualized ecological momentary interventions to promote engagement in and maintenance of healthy behaviors. Achieving this goal requires high-resolution data on multiple time scales of observation and at multiple levels of analysis (e.g., environmental, social, behavioral, physiological, genetic).

To Attend

Visitors to RAND's Santa Monica and Pittsburgh locations are welcome to attend & must RSVP at least one day prior to the seminar. To ensure attendance please, contact Donna Mead with your name, company or affiliation & national citizenship (for security purposes).

Sponsored by the RAND Center for Causal Inference (CCI) and the RAND Statistics Group