Disputes About Statistical Generalizations

Abstract

Discussing differences between samples, remember that within large samples, even a small difference might be statistically significant yet not necessarily interesting or relevant to the concern that has prompted it. So you may still want to inquire whether any given difference is worth discussing. Prior to debating assertions that explain certain event, investigate first the assertion that the event in question is due to a mere accident. Avoid debates on causality (unless your concern is philosophical). Discussing the degree to which the appearance of one variable makes us expect the appearance of another usually suffices.