Statistical Dogma: A Challenge
Irwin D. J. Bross · The American Statistician · 1961
In the long run, I think, it is disastrous for statisticians to be dogmatic. Yet modern statistics (especially the theoretic phases) is often dogmatically presented. This may be expedient (eg. confer a short term advantage) in persuading some students or customers that statistical methods are a marvelous invention (and, incidentally-, that the person who makes these forceful and sweeping assertions really knows his stuff). But when the reaction does comeand it has come in areas where statistics was oversold-it is equally forceful and sweeping. The pendulum swings of dogmas also disrupt the progress of statistics. Due to Laplace, Bayesian theories acquired the force of During the Fisherian reaction Bayes theorem virtually disappeared from statistics texts. Fisher's vigorous and total repudiation is the usual counterreaction to In the past decade the pendulum has swung again and the new Bayesians are as dogmatic as the old ones. Let me make clear what I mean by a dogma. To rue a dogma is a statement (or conceptual system) for which: