causal inference · Hybrid
Matching and Propensity Scores
Overlap common support, balance tables, and sensitivity analyses for observational comparisons.
ZAR 1 120 · 5 weeks
Software: Stata · Method focus: matching · Level: intermediate · Assessment: balance report
Description
Implement nearest neighbor and kernel matching with honest reporting of remaining bias risks. Includes love plots and standardized difference practice.
What is included
- Overlap trimming exercises
- Balance table automation
- Sensitivity discussion prompts
- Mentor review of love plots
- Stata and R examples
- Short reading on doubly robust methods
Outcomes
- Diagnose bad overlap before matching
- Present balance with transparent thresholds
- Discuss sensitivity without claiming elimination of bias
Dr. Zanele Khumalo
Econometrics lecturer focused on observational causal methods.
Reviews
Love plot feedback was specific — we finally standardized covariates consistently.
Overlap trimming exercise prevented a headline that would not have survived audit.
Questions
Conceptual overview; full DR programming is advanced follow-up.
Not covered — we stick to propensity and distance matching basics.
Mentioned as pointer only.