Propensity Score Analysis in Multilevel Settings

• Activity Number: 475
• Type: Invited
• Date/Time: Wednesday, August 6, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Social Statistics Section
• Author(s): Aarti P. Bellara*+ and Eun Sook Kim
• Companies: University of South Florida and University of South Florida
• Keywords: Propensity Score Methods ; Multilevel Models ; Observational Studies

Abstract

Propensity score (PS) analysis is a recently popular method for reducing the selection bias in observation studies to identify causal relationships. Conducting a propensity score analysis is a multi-step decision based process that generally includes selecting the covariates, estimating the PS model, evaluating the estimated PSs with the possibility of revising the PS model, and finally applying the estimated PSs to an outcome model. Several researchers have outlined these steps and provided general guidelines that have helped scholars demystify the process of using propensity scores. The application of propensity score methods with hierarchical or nested data has not received as much attention. This paper aims to outline the steps for conducting a propensity score analysis within a hierarchical framework including the similarities and differences with single-level propensity score methods. A recent study that evaluated the effectiveness of using propensity score methods with nested data will be used as an example throughout the paper.