Statistics Done Wrong: The Woefully Complete Guide

by Alex Reinhart

leave-one-out cross-validation, where the model is fit using all but one data point and then evaluated on its ability to predict that point;