Bayes's Theorem: Response Scales and Feedback.
Abstract
Six groups of subjects provided a test of the relative effectiveness of various procedures for using Bayes's theorem to revise probabilites in the light of new information. The task was to predict whether Port A or Port B was the destination of a hypothetical ship on the basis of information about the age of the ship, gross tonnage, percent of cargo capacity on board, and amount of fuel taken on. The historical data were normally distributed with different means for each port. The use of log scales (either odds or likelihood ratio) led to a deterioration in performance as compared with no scales, except when the log odds scale contained a display of associated posterior probabilities. The probability scale apparently constained posterior probabilities from becoming overly extreme. In the absence of scales, likelihood ratio assessment resulted in more nearly optimal performance than did the verbal assessment of posterior odds; but the use of posterior odds feedback slightly improved the assessed likelihood ratios. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 30, 1972
- Accession Number
- AD0751563
Entities
People
- Barbara C. Goodman
- Cameron R. Peterson
- Patricia A. Domas
Organizations
- University of Michigan