The Trinity of Statistics (Departmental Seminar)

Abstract:

There are three schools of thought in statistics: frequentism, Bayesianism, and likelihoodism. They are often thought to be in fundamental disagreement, but I don’t think so. My goal is to develop a unification of the three camps, and resolve the most urgent of the conflicts among them: (i) Lindley’s paradox, (ii) the optional stopping problem, (iii) the Neyman-Scott and James-Stein paradoxes, and (iv) the debate on actualism. My focus on statistics might appear to be narrow, but the results are meant to be wide-ranging and should interest at least four groups of people. (1) For philosophers of science, I point out a way to a more unified account of scientific inference than supposed before. (2) For historians of science, I identify a movement within the statistical practice that seeks unification rather than confrontation—a movement underway since at least the 1960’s but yet to be more recognized outside of statistics. (3) For formal epistemologists, I make explicit a way to unify the epistemology of credences and that of all-or-nothing beliefs, just using Bayesianism and classical frequentism as examples. (4) For traditional epistemologists, I offer a hope for an internalist guide to the reliabilist epistemology of inference. Or so I argue.

Delivered in English.
All are welcome.