Propensity Score Analysis

A 2-Day Seminar Taught by Shenyang Guo, Ph.D. 

Propensity score analysis is a relatively new and innovative class of statistical methods that has proven useful for evaluating the effects of treatments or interventions when using nonexperimental or observational data.

Although regression analysis is most often used to adjust for potentially confounding variables, propensity score analysis is an attractive alternative. Results produced by propensity score methods are typically easier to communicate to lay audiences. And propensity score estimates are often more robust to differences in the distributions of the confounding variables across the groups being compared.

This seminar will focus on four closely related but technically distinct propensity score methods:

  • Heckman’s sample selection model and its revised version for estimating treatment effects using Stata treatreg
  • Propensity score matching and related methods, including greedy matching, optimal matching, and propensity score weighting using Stata psmatch2, pweights and R optmatch
  • Matching estimators using Stata nnmatch
  • Propensity score analysis with nonparametric regression using Stata psmatch2 and lowess.

The examination of these methods will be guided by two conceptual frameworks: the Neyman-Rubin counterfactual framework and the Heckman scientific model of causality. The course also covers Rosenbaum’s approaches of sensitivity analysis to discern bias produced by hidden selections.

The seminar uses Stata software to demonstrate the implementation of propensity score analysis. Although participants will not do hands-on work during the seminar, they are encouraged to practice on their own time. All syntax files and illustrative data can be downloaded at the Propensity Score Analysis Support Site.

Who should attend

The seminar will be helpful to researchers who are engaged in intervention research, program evaluation, or more generally causal inference, when their data were not generated by a randomized clinical trial. The prerequisite for taking this seminar is knowledge of multiple regression analysis. Researchers from economics, public health, epidemiology, psychology, sociology, social work, medical research, education, and similar disciplines may consider participating.

Location and Materials 

The course meets 9 a.m. to 4 p.m. on Friday, October 12 and Saturday, October 13 at Temple University Center City, 1515 Market Street, Philadelphia, PA.

Participants receive a bound manual containing detailed lecture notes (with equations and graphics), examples of computer printout, and many other useful features. This book frees participants from the distracting task of note taking. 

Registration and lodging

The fee of $895.00 includes all seminar materials. 

Lodging Reservation Instructions

A block of guest rooms has been reserved at the Club Quarters Hotel, 1628 Chestnut Street, Philadelphia, PA at a special rate of $137 per night for a Standard room. This location is about a 5 minute walk to the seminar location. In order to make reservations, call 203-905-2100 during business hours and identify yourself by using group code STA111. For guaranteed rate and availability, you must reserve your room no later than September 11, 2012.

Seminar outline

  1. Observational studies and challenges
  2. Why and when propensity score analysis is needed
  3. The Neyman-Rubin counterfactual framework
  4. The assumption of strongly ignorable treatment assignment
  5. The stable unit treatment value assumption
  6. Truncation, censoring, and incidental truncation
  7. Key features of Heckman’s sample selection model
  8. Key features of treatment effect model; Stata-treatreg
  9. The Rosenbaum and Rubin model (1983)
  10. Greedy matching; Stata-psmatch2
  11. Optimal matching; R-optmatch
  12. Post-optimal-matching analysis
  13. Propensity score weighting
  14. Matching estimators; Stata-nnmatch
  15. The kernel-based matching estimator; Stata-psmatch2
  16. Rosenbaum’s sensitivity analysis; Stata-rbounds
  17. Overview of Heckman’s model of causality 
  18. Criticism of nonexperimental approaches
  19. Directions of future development