Now published in final form: our commentary pleading for methodology over metrics. Open access in Journal of Clinical Epidemiology 👉 bit.ly/3BQgtga

Whether I'm writing trading code or simulations or research infrastructure, I've almost never regretted taking the path of simplicity/less generality/fewer abstractions. Can add that stuff later if needed. All my software regrets come from over-engineering at the beginning.

Them: "Can you help us do sample size calculation?" Me: “OK” Them: “Oh, that comes out much bigger than we can afford so we will just do what we planned to do anyway.”

Would be fun to see the responses to this if one used it as an interview question.