Introduction Suppose you have reason to believe that a biomarker signal could precede a clinical diagnosis, leading to earlier and better intervention. You have a large collection of candidate biomarkers, and you think possibly a small number of them could form the basis of this predictive signal. You assemble a collection of clinical subjects for…

This essay goes out to statisticians who are wondering how Machine Learning (ML) compares to classical or parametric statistics, whether it will affect their jobs, or what role it should have in their current work. I recently presented some thoughts at the International Alliance for Biological Standardization’s annual statistics conference; the audience was primarily statisticians…