Doing thousands of hypothesis tests at the same time

摘要

The classical theory of hypothesis testing was fashioned for a scientific world of single inferences, small data sets, and slow computation. Exciting advances in scientific technology - microarrays, imaging devices, spectroscopy - have changed the equation. This article concerns the simultaneous testing of thousands of related situations. It reviews an empirical Bayes approach, encompassing both size and power calculations, based on Benjamini and Hochberg's False Discovery Rate criterion. The discussion precedes mainly by example, with technical details deferred to the references.