Hacker News new | ask | show | jobs
by timy2shoes 720 days ago
Then take a Cauchy or a t-distribution. Basically anything with a longer tail than exp(x^2). The Gaussian summary will be misleading because of the tails.