False discovery rate

Ever wonder how scientists avoid falsely claiming new discoveries?

Imagine you're tasting different wines at a party. You want to say which ones are good, but you don't want to mistakenly praise a bad wine as good.

Think of a method that helps you say "good wine" without too many mistakes, even if you're tasting many wines. It's like a balance between being careful and being confident.

Example

Out of 100 wines tasted, you confidently say 10 are good. But with a strict rule, you might only say 2 are good to avoid mistakes. The Benjamini-Hochberg procedure helps find a middle ground, allowing you to confidently say 7 are good while keeping mistakes low.

Remember this

The Benjamini-Hochberg procedure helps you confidently identify good wines while minimizing the chance of mistakenly praising a bad one.

Text adapted from Wikipedia, licensed under CC BY-SA 4.0.

Related concepts

Clinical trial

Why do scientists trust their experiments?

Kolmogorov–Smirnov test

Ever wondered how scientists know if two things are the same or not?

Adam has bias correction: divides by (1-β^t) in early steps

Why do we sometimes need to fix mistakes in computer decisions?

the Bonferroni correction does: divides α by number of tests

Bonferroni correction adjusts significance level by dividing α by the number of tests

Manifold hypothesis

High-dimensional data lies on lower-dimensional manifolds

Overlapping subproblems

Ever calculated a huge Fibonacci sequence by hand?

Swipe through more Machine Learning concepts

Open Pocket Polymath