Union-of-intersections test sequences (UI-TSs)

Code for Spertus, Sridhar, and Stark (2026+). Contains tools to run hypothesis tests about means of bounded, stratified populations with sequential, finite-sample, nonparametric (SFSNP) level control. The primary tools are union-of-intersections test sequences (UI-TSs) and sequentially valid lower confidence bounds (LCBs).

The Code directory contains the following python scripts:

• utils.py contains all functions implementing the proposed methods
• test.py contains unit test for all functions
• significance_simulations.py runs the t-test simulation that generated Figure 1
• pointmass_simulations.py runs the pointmass simulations presented in Section 7.1
• bernoulli_simulations.py runs the Bernoulli simulations presented in Section 7.2
• gaussian_simulations.py runs Gaussian simulations (not currently described in the manuscript)
• pgd_time.py times the convex UI-TS against the LCB method, as described in Section C.4
• r_plots.R contains the R code used to generate all plots in the paper

The Results directory contains CSV files with the simulation results