Permutation Tests

An Interactive Introduction

0.1 Welcome

This lecture note introduces permutation tests — an exact, distribution-free approach to hypothesis testing that does not rely on asymptotic approximations or parametric assumptions.

The central idea is simple: under the null hypothesis of no treatment effect, the observed outcome vector \(Y\) is independent of the treatment assignment \(W\). If that is true, then any re-assignment of \(W\) is equally likely, and we can construct the exact null distribution of any test statistic by enumerating all such re-assignments.

0.2 What You Will Find Here

Chapter	Content
2 — Permutation Tests	Setup, test statistic, permutation distribution, critical values, and an interactive simulation

0.3 Setup

library(tidyverse)