A couple weeks ago I took a great 2-day stats course on evaluating treatment effects using observational data that was offered by Statistical Horizons and taught by Dr. Stephen Vaisey, Professor of Sociology at Duke. Terrific stuff. Learned key concepts about new and better ways to do such analyses. One gem that my colleague, Tim Hediger, recommended and which resonated with the theme of the course was this YouTube video by Dr. Gary King at Harvard click here. It clearly lays out why, despite its buzz … propensity score matching is NOT the best way to match observational data prior to analyzing treatment effects.
Fellow geeks, enjoy!