"Some comments on the history of statistics and the evaluation of effect size"
Statistics is typically taught to psychologists as unquestionable truth akin to mathematics, rather than as the philosophically complex and controversial topic that it is. Neglected issues include the meaning(s) of “probability,” the paradoxes that arise when rational judgment is defined in terms of statistics, the quest for numerical tools to make substantive decisions for us, and most of all, the fraught history and meanings of the term “significance.” I plan to briefly survey some of these topics, hoping to inspire listeners to read and learn more on their own, and then add some brief remarks on the evaluation of effect size. As a special case of the neglected issues in statistics, effect sizes have been simplistically and mindlessly evaluated, with special problems arising from common practices such as squaring correlations to yield “percentage of variance explained,” and the use of customary, arbitrary, and fundamentally meaningless standards. Instead, effect sizes should be evaluated in comparison to well-understood benchmarks or in terms of concrete consequences.