Propensity Score Analysis: A Friendly Tutorial

Social work researchers often wish to assess intervention effectiveness by comparing two interventions or by comparing a single intervention to usual care. When ethical considerations or other factors preclude the use of an experimental design, the researcher must utilize a quasi-experimental design. In a quasi-experimental study, uncontrolled pre-intervention differences between groups can result in a biased estimate of the intervention effect. Propensity score analysis provides a way of balancing treatment groups on measured covariates in order to remove bias. In this workshop, the presenter will provide a gentle introduction to propensity score analysis and will describe best practices.

The workshop is designed to be interactive, intuitive, and light on math. In order to tailor the workshop for participants, the presenter will briefly assess participants’ experiences and goals in relation to propensity score analysis. The workshop will cover four main content areas: (1) definition and basic theory of the propensity score, (2) differences between propensity score analysis and conventional regression modeling, (3) propensity scoring assumptions, and (4) steps of propensity score analysis. The analysis steps represent the primary focus of the workshop. We will discuss covariate selection, propensity score estimation, propensity score application methods, and a number of assumption checks, model adjustments, and sensitivity analyses designed to maximize bias reduction and support excellence in analysis and reporting. The workshop will include ample opportunities for questions and discussion.