Mariel Finucane from Mathematica
Abstract: It’s an old trick among statisticians: when we can’t answer the question that our collaborators are substantively interested in, we answer a question that sounds like that question. We hope they won’t notice the bait and switch. That, in a nutshell, is the story of statistical significance in the field of social policy evaluation. People want some assurance that the policy being studied moved the needle for the individuals or communities it serves. Statistical significance and p-values sound like they can provide that assurance. They can’t, and when we pretend they can (which we still do horrifyingly often!), we can be making a large-magnitude – not just semantic – mistake. In this talk, we provide examples of how statistical significance is an obstacle to achieving our mission, and we describe how Bayesian methods using evidence-based priors can give us what we thought we were getting with statistical significance, and more.
Mariel will be on-site for the seminar in W2008, but the seminar will also be streamed through Zoom.
Meeting ID: 928 8613 3715