On the choice of a consensus distribution in Bayesian analysis
Four methods of arriving at a consensus distribution in Bayesian analysis are described (weighted-average, natural-conjugate, feedback and reassessment, and group reassessment). The weighted-average and natural-conjugate methods are compared empirically in probability choice and generation models. The use of the weighted-average method was more frequent and its use increased in the presence of high variance among experts' judgments. The cognitive abilities, tolerances for ambiguity, and risk-taking propensities of the subjects were not related to their choices of a consensus distribution. Implications for future research are drawn. © 1972.
Winkler, RL; Cummings, LL
Volume / Issue
Start / End Page
International Standard Serial Number (ISSN)
Digital Object Identifier (DOI)