Short Course: Causal Inference and Treatment Effect Estimation Using Stata

Causal Inference and Treatment Effect Estimation Using Stata

Enrique Pinzon
Full Day, 9:00 AM – 5:00 PM
Palais des congrés de Montréal, 512D

In this workshop, we discuss methods for drawing causal inferences when analyzing observational rather than experimental data. We present a variety of
estimators for average treatment effects (ATEs) and average treatment effects on the treated (ATETs) and discuss when each estimator is useful. Throughout the workshop, we cover the conceptual and theoretical underpinnings of treatment effects and demonstrate the methods with many practical examples worked using Stata software.

After a discussion of the potential-outcome framework and an overview of the parameters estimated, the workshop introduces the following treatment-effect estimators

  • regression-adjustment estimator
  • inverse-probability-weighted (IPW) estimator
  • augmented IPW estimator
  • IPW regression-adjustment estimator
  • nearest-neighbor matching estimator
  • propensity-score matching estimator
  • difference-in-differences (DID)

The course also discusses

  • standard errors and diagnostics for DID estimation
  • double-robustness property of the augmented IPW and IPW regression-adjustment
  • estimators using different functional forms for outcome model and treatment
  • model multivalued treatments
  • estimators when the treatment is endogenous

The discussion of estimators that handle an endogenously assigned treatment includes extended regression model (ERM) estimators, which can also
account for other complications in observational data such as endogenous sample selection and endogenous regressors.

All topics are discussed using a combination of theory and Stata examples.