La calibrazione fiduciale nella sintesi di opinioni "esperte": una proposta di stima delle varianze degli indicatori in performance

Abstract

In conditions of uncertainty, information regarding an unknown entity (a physical quantity, an event, a risk) can derive from multiple sources. Hence, the coherent synthesis of heterogeneous sources of knowledge, often with different informative weight, must pass through a Bayesian composition of the available information. With reference to Peter A. Morris' Bayesian aggregative algorithm (1974, 1977, 1983, 1986) and a fiducial modelisation of the calibration function for the experts' opinions (Monari e Agati, 2001), this paper addresses the problem of the performance indicator variances assessment: the proposed solution is based on the Delta method, that uses the Taylor formula to make derivable functions linear.