Paper for EPSA 2015 proceedings

Generalized confirmation and relevance measures


Abstract. The main point of the paper is to show how popular probabilistic measures of incremental confirmation and statistical relevance with qualitatively different features can be embedded smoothly in generalized parametric families. In particular, I will show that the probability difference, log probability ratio, log likelihood ratio, odds difference, so-called improbability difference, and Gaifman’s measures of confirmation can all be subsumed within a convenient biparametric continuum. One intermediate step of this project may have interest on its own, as it provides a unified representation of graded belief of which both probabilities and odds are special cases.


Preprint available HERE.


Questa voce è stata pubblicata in Altro. Contrassegna il permalink.

I commenti sono chiusi.