Intro

I don’t know why, but back in high school, hypothesis testing just didn’t click the first time I learned it. Freshman year of college wasn’t much better, honestly. The mechanics were fine, reject if p is small enough, but the why behind any of it stayed fuzzy for a long time.

This book is my attempt to fix that for myself, in public. Starting from the hypothesis testing everyone learns first, working through what I’ve picked up in courses since, and pushing out toward papers I’m reading right now.

It’s a working notebook, not a finished textbook. Some chapters below are written out. Others are still just an outline I’m filling in as I go, treat anything marked TODO exactly as that.

Where this is going Before breaking anything, I’ll quickly review the classical frequentist testing, so that we can have the standard picture on the table first. After that, this book pulls at each assumption baked into that picture one at a time: extending it toward Bayesian and likelihood-based tests, then toward testing many hypotheses at once, and finally toward where any of this actually shows up inside machine learning.