### Department of Biostatistics Seminar/Workshop Series

# Propensity Scores: Recent Developments

## Patrick Arbogast, PhD

### Associate Professor, Department of Biostatistics

Vanderbilt University School of Medicine

### Wednesday, December 3, 1:45-2:55pm, MRBIII Conference Room 1220

### Intended Audience: Persons interested in applied statistics, statistical theory, epidemiology, health services research, clinical trials methodology, statistical computing, statistical graphics, R users or potential users

Propensity scores are an increasing popular methodology for adjusting for a large number of confounders. I have recently learned of some papers regarding the application of propensity scores that will influence its use in the future. I will talk about two of these topics. First, for a binary exposure, logistic regression is typically used to estimate propensity scores. However, due to the non-collapsibility of the logistic model, one may get biased exposure effect estimates. However, if the log-linear link is used in the regression model relating outcome to exposure, exposure effect estimates will be unbiased. My second topic regards variable selection for propensity score models. A recent paper demonstrated how including covariates in the propensity score model that are related to the outcome or exposure, but not both, can affect bias, variance, and MSE’s of exposure effect estimates. I will summarize these papers and present some simulation results.