A deep-dive companion to the experimentation refresher, expanding the three methods covered as a tour in that course's Module 8: modern difference-in-differences, synthetic control, and causal forests for heterogeneous treatment effects.
| # | Module | Status |
|---|---|---|
| 1 | TWFE Diagnosed — Goodman-Bacon decomposition, the zoo of 2×2s, when TWFE is unbiased | ✓ done |
| 2 | Heterogeneity-Robust DiD — CS, SA, BJS, dCDH in detail | ✓ done |
| 3 | Honest DiD — sensitivity bounds for parallel trends | ✓ done |
| 4 | Synthetic Control — estimator, inference, augmented and generalized variants | ✓ done |
| 5 | Synthetic DiD — the bridge from SC to DiD | ✓ done |
| 6 | Causal Forest — honest splitting, asymptotic normality, GRF | ✓ done |
| 7 | Policy Learning — from τ̂(x) to deployment rules | ✓ done |
| 8 | Matrix Completion — the modern panel toolbox, surrogate index, and the capstone decision tree | ✓ done |
R: fixest, did, didimputation,
DIDmultiplegt, HonestDiD, Synth,
gsynth, synthdid, augsynth,
grf, plus tidyverse.
See the learning plan for the module-by-module spec, or the README for written concept references per module.