UPE Seminar: Stuart Hurlbert
Abstract: As practiced by most scientists, statistics is first and foremost a religion with mathematical trappings that function to scare the intuitive, the rebellious, and the heretical into submission. What is done and taught with respect to particular statistical issues shows little consistency from one nation to another, one university to another, one textbook to another, one discipline to another, one editor to another. Where there is some consistency, the majority is wrong. These problems have their main origin in polemics and cryptic writing of a few brilliant gents in England in the early part of the last century. Peeling away inherited doctrinal pathologies, we discover one should never specify alpha, never use the word significant, never admit to the possibility of type II error, never get paranoid about multiple comparison situations, never use one-tailed tests, and never use repeated measures analysis of variance. For sinners, an opportunity for absolution will be offered at the end of the talk.