Regression and Matching (3).pdfEconometrics, Statistics, Financial Data Modeling (2017)
In Mostly Harmless Econometrics Angrist and Pischke discuss regression in the context of matching. Specifically they show that regression provides variance based weighted average of covariate specific differences in outcomes between treatment and control groups. Matching gives us a weighted average difference in treatment and control outcomes weighted by the empirical distribution of covariates. These notes sketch out the argument using potential outcomes and expectations notation.
- causal inference
Publication DateJuly, 2017
Citation InformationMatt Bogard. "Regression and Matching (3).pdf" Econometrics, Statistics, Financial Data Modeling (2017)
Available at: http://works.bepress.com/matt_bogard/37/