The Reliability of Subjective Probabilities Obtained Through Decomposition

The use of decomposition as a procedure for improving the consistency of subjective probability encoding is discussed. Using a psychometric measurement model, an expression is developed that describes the random error associated with decomposition estimates as a function of characteristics of the component assessments. Decomposition is compared to direct assessment in terms of the percent change in measurement error that can be attributed to the use of decomposition. Potential benefits of decomposition are specified and recommendations made on how to utilize decomposition as an approach for error control.

• 저널Management Science · 34(2) · 186–199
• 토픽Bayesian Modeling and Causal Inference · Artificial Intelligence
• DOI10.1287/mnsc.34.2.186
• 저자H. V. Ravinder, Don N. Kleinmuntz, James S. Dyer