Introduction to matching and propensity score analysis

Home / Research Methods  / Introduction to matching and propensity score analysis

A 6-hour workshop taught by Steve Porter, Ph.D.

You may also be interested in our introduction to binary logistic regression workshop.

See when this workshop is offered next »

Overview of our matching and propensity score analysis workshop

Propensity score analysis (also known as “matching”) is a popular way to estimate the effects of programs and policies on outcomes. Yet researchers face a dizzying array of choices, in terms of particular matching techniques to use, as well as many different options for implementing a specific technique.

This workshop provides a concise introduction to matching for the applied researcher. Rather than cover every possible matching technique, we will focus on nearest neighbor matching (one of the most popular approaches) and inverse propensity weighting, a simple and powerful matching approach that can be used without any specialized software (some software packages, like SAS and SPSS, do not come with built-in matching commands, requiring the use of often opaque and difficult to use macros).

Expected outcomes

By the end of the workshop, participants should understand why matching is preferred over regression, the major concepts underlying the counterfactual theory of causality, the major issues with implementing nearest neighbor matching, and whether they should estimate the average treatment effect or treatment effect for the treated for their particular research application. Most importantly, they should be able to immediately begin using inverse propensity weighting in their research, using any statistical software program.

Who should attend?

The target audience is researchers who typically use multiple regression, logistic regression and hierarchical linear modeling in their research and 1) wish to know why matching has become popular, and 2) how to use matching in their research. Participants should have a good understanding of multiple regression. Familiarity with logistic regression is helpful but not required. If you want to learn about logistic regression, consider our workshop on logistic regression.

Agenda

  1. Advantages of matching over regression and other linear models
  2. Rubin’s Causal Model – the counterfactual approach to causality
  3. Understanding treatment effects for research projects
  4. Implementing nearest neighbor matching by hand and with psmatch2
  5. Choosing the right variables for the propensity model
  6. Advantages of inverse propensity weighting over all other matching techniques
  7. Implementing inverse propensity weighting using survey weight commands
  8. Common mistakes to avoid when matching

See when this workshop is offered next »