Inspired by the lack of blog posts about multiple hypothesis testing, I wrote a post on multiple hypothesis testing for the Multithreaded blog. The post covers three aspects of hypothesis testing:

• The difference between p-values and the “probability of being right”
• How to combine many hypotheses into a single test
• How to deal with multiple hypotheses individually

With the help of the Algo UI team, it also contains interactive visualization that showcase the differences between frequentist and Bayesian viewpoints on testing, and how the Benjamini-Hochberg procedure controls the false discovery rate.

The material is based on the fantastic Stanford Statistics 300C course taught by Prof. Candes.