Modeling Expert Opinion: Likelihoods under Incomplete Probabilistic Specification
Abstract
Expert opinion is often sought with regard to unknowns in a decision-making setting. The presumption is that such opinion is elicited as an incomplete probabilistic specification either in the form of probability assignments to fixed intervals or in the form of selected quantiles. The authors present likelihoods for such a specification that arise through random mixtures of Beta distributions. They presume that a supra Bayesian presides over the opinion collection resulting in the posterior distribution as the mechanism for pooling opinion. The models are applied to opinions collected regarding points per game for participants in the 1991 NBA championship basketball series.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 09, 1992
- Accession Number
- ADA258775
Entities
People
- Alan E. Gelfand
- Bani K. Mallick
- Dipak K. Dey
Organizations
- Stanford University