Early Steps Towards Generalizability Equivalence (with implications for adaptive design)

Google map

Abstract: 

Generalizability is an important topic in the behavioral sciences, but is often used without a precise definition, or with definitions not suited to comparison of designs or scientific approaches. In this talk, I will present early-stage work in collaboration with colleagues at the Max Planck Institute in Berlin in which we explore generalizability in terms of what a published result tells us about the world. I discuss the idea of this formalization and its uses to compare scientific approaches, and illustrate via simulation and math a few of its unintuitive implications with regard to design and analysis decisions.