Sign-flip randomization test

A sign-flip randomization test independently multiplies observations by signs in $\{-1,1\}$. It is finite-sample exact for a zero-centered null when the joint distribution is invariant under those sign changes.