Regression Adjustment and Stratification by Propensity Score in Treatment Effect Estimation

• Activity Number: 402
• Type: Topic Contributed
• Date/Time: Tuesday, August 3, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Health Policy Statistics Section
• Author(s): Jessica Amelia Myers*+ and Thomas Louis
• Companies: Johns Hopkins Bloomberg School of Public Health and Johns Hopkins Bloomberg School of Public Health
• Keywords: Generalized Additive Model ; Observational study ; Optimal strati cation ; Propensity score adjustment

Abstract

Propensity score adjustment of effect estimates in observational studies of treatment is a technique used to control for bias in treatment assignment. Regression adjustment is used most often, but it can lead to biased results when model assumptions are violated. Validity of the stratification approach depends on fewer model assumptions, but is less efficient when the regression assumptions hold. We compare two stratification approaches, equal frequency classes and an approach that attempts to minimize the mean squared error (MSE) of the treatment effect estimate, with a Generalized Additive Model (GAM) regression approach, that flexibly estimates the relations among propensity score, treatment, and outcome. We find that under many data generating distributions the GAM approach outperforms stratification in treatment effect estimation. We illustrate approaches via an example analysis.