Calibration-based empirical probability (Q1072208)

The paper gives a general theory of empirical probability relative to an informal base. Since this theory does not require that probabilities be interpreted in terms of repeated trials of the same event, it can be seen as a new generalization of traditional frequentist interpretations of probability. The main result of the investigation shows that any two sequences of forecasts which both meet a natural criterion of empirical validity must be in asymptotic agreement. Finally one can find a commentary of M. J. Schervish on the main theorems, and a short answer of the author.