CONCERNING THE SIMULATION OF DIAGNOSTIC SYSTEMS WHICH PROCESS COMPLEX PROBABILISTIC EVIDENCE SETS.
Abstract
The report provides a theoretical and methodological background for a series of experiments being performed on diagnostic systems. The basic methodology underlying current simulation of diagnostic environments has been changed considerably to achieve the level of task complexity required for the study of certain problems judged to be of significance in real diagnostic systems. In the current series of experiments, the Bayesian evidence aggregation or PIP procedure will be compared with the procedure of having men make direct estimates of posterior probabilities from data. The latter procedure will be referred to as POP. The major reason for continuing the comparison between these two procedures is that results from the POP procedure will provide important additional base-line data in the evaluation of the PIP procedure. Although one can now expect PIP to be generally superior to POP, it will often be of considerable interest to determine how much superior PIP is in various experimental circumstances. The current simulation methodology allows for several theoretical standards for evaluating the accuracy of posterior probabilities produced by the PIP and POP procedures. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 01, 1969
- Accession Number
- AD0691238
Entities
People
- David A. Schum
Organizations
- Ohio State University