An R package on CRAN and Github of two sample tests, testing the null that two samples come from the same distribution.

• DTS: novel statistic proposed in associated paper here (and on arXiv)
  • Approximately the integral of variance weighted ECDF differences.
  • High power in variety of settings. (See comparisons to maximal power (T-test and F-test) in the paper)
• Other included test statistics: Kolmogorov-Smirnov, Kuiper, Cramer-Von Mises, Anderson-Darling, Wasserstein
• Fast (C++) test statistic calculations
  • Critical for slower ECDF based tests
• Randomization for the null distribution and p-values
  • Validity under assumption that observations are independent

Synthetic Controls With Spillovers

R code implementing the routine proposed in my paper with Jianfei Cao.

• Calculate SCM weights for every observation in the data
• Estimates treatment effect with spillovers
• Inference based on ‘random treatment timing’
• Data and code to run cigarrette example in our paper
• See Jianfei’s website for Matlab code.