Roadmap

[nignatiadis]

Roadmap

Approaches to inference in sharp RDs

• [ ] Robust bias-corrected inference in rdrobust package

• [x] Noise Induced Randomization (todo: document)

• [ ] Inference accounting for worst-case smoothness

  • [x] OptRDD (Imbens and Wager, 2019)
  • [ ] RDHonest
  • [ ] Bounded misspecification inference for discrete RVs (https://www.aeaweb.org/articles?id=10.1257/aer.20160945)

• [ ] Local randomization rdlocrand

• [ ] Naive local polynomial regression (not really inference..)

  • [x] p = 1 (local linear)
  • [ ] p > 1

Data-driven parameter choice

Bandwidths

• [ ] Imbens-Kalyanaraman bandwidth
  • [x] p = 1 (local linear)
  • [ ] p > 1
• [ ] RDrobust second generation bandwidths for point and bias estimation

Curvature bounds

• [ ] RDHonest heuristic

Manipulation

• [x] McCrary density test
• [ ] Density tests in the rddensity package.
• [ ] Partial identification results under manipulation

Plotting

• [ ] Histogram of running variable
• [ ] Basic binscatter
• [ ] RD plots and binsize selection as in (https://rdpackages.github.io/references/Calonico-Cattaneo-Titiunik_2015_JASA.pdf)

Empirical sanity checks

• [ ] Sensitivity analysis
• [ ] Placebo tests

Longer term extensions

• [ ] Fuzzy designs
• [ ] Multidimensional running variable
• [ ] RD with covariates