Probability and Statistics II

The notion that imprecise credences are rationally permissible or even mandatory has gained increasing attention. Less well known is that there are powerful arguments that chances are sometimes imprecise. I propose a specific principle for how rational credence would calibrate to imprecise chance. The principle entails that imprecision of (known) chances begets imprecision of rational credence. Read one way this principle, together with the assumption that there exist imprecise chances, provides strong additional reason for thinking that rational credence is sometimes imprecise. Read another way, it provides a criterion for evaluating claims that certain sets of probabilities constitute imprecise chances.

This paper investigates the conceptual relationship between openness and reproducibility using a model-centric approach. We first clarify the concepts of reliability, auditability, replicability, and reproducibility--each of which denotes a potential scientific objective. Using the notion of an idealized experiment, we identify which components of an experiment need to be reported and which need to be repeated to achieve the relevant objective. The model-centric framework we propose aims to contribute precision and clarity to the discussions surrounding the so-called reproducibility crisis.

Pursuit of predictive accuracy is often said to have a paradigmatic example in model selection, using the technique developed by Akaike and other statisticians (Forster and Sober 1994). I like pursuit of predictive accuracy, but I raise seven worries about the Akaike-style approach to predictive accuracy.