|Friday, February 16|
|CS10 Propensity Scores and Resampling Methods||
Fri, Feb 16, 2:00 PM - 3:30 PM
CANCELED: A Streamlined Process for Conducting a Propensity Score-Based Analysis (303551)*John A. Craycroft, University of Louisville
Maiying Kong, University of Louisville
Keywords: propensity score, observational data, treatment effect, causal inference, unbiased
Propensity scores are used in a large class of methodologies in order to balance covariate distributions among treatment groups in order to obtain unbiased estimates of treatment effects. The typical process involves several major decision points, including which covariates to use in estimating the propensity scores. This presentation will: give background and context for propensity score methodology; summarize conclusions from recent literature regarding which covariates to include in the propensity score model; describe a user-friendly algorithm for processing any data set, even with p>>n, and for developing propensity scores in the recommended manner; demonstrate the algorithm using R. This process can be employed in any observational data setting to efficiently conduct a propensity score analysis.