Riffyn Blog

If you’re confused about the differences between machine learning, AI, deep learning, supervised learning, or unsupervised learning, you’re not alone. This blog explains the differences, and reveals that you may have even been doing machine learning for years without even knowing it.

As a full-time scientist you test a lot of hypotheses — a lot more than one per week. If you test tens or hundreds or even thousands of samples in a single experiment, it’s guaranteed that you’ll find many false positive results, in every experiment. So what can you do about it?

FDR is a very simple concept. It is the number of false discoveries in an experiment divided by total number of discoveries in that experiment. But there is a problem, you never know how many discoveries are actually real or false when you accepted them. So how do you estimate FDR from your data?

Family-Wise Error Rate (FWER) is an approach for multiple testing correction. It is a scary sounding term, but it’s simply the probability that one or more of your “family” of multiple tests is false. FWER offers a generally more stringent approach than FDR to reduce your false findings.