Module stats 

Non-parametric stats for experiment reports.

Effect size = median(treatment) − median(control). CI via percentile bootstrap (default 10k resamples, 95%). Winsorize p1/p99 before resampling to blunt skew.

Re-exports

pub use bootstrap::winsorize;

pub use srm::has_srm;

Modules

bootstrap

Bootstrap CI and winsorization helpers.

cuped

CUPED — Controlled-experiment Using Pre-Experiment Data.

power

Minimum Detectable Effect (MDE) at 80% power / 95% CI.

sequential

Sequential / always-valid testing via the mixture mSPRT approach.

srm

Sample Ratio Mismatch (SRM) detection.

Structs

Summary

Constants

DEFAULT_RESAMPLES

MIN_SAMPLE

Functions

summarize

Pure stats for a metric. Deterministic given seed.