NCJ Number
86504
Journal
Journal of Police Science and Administration Volume: 10 Issue: 3 Dated: (September 1982) Pages: 335-343
Date Published
1982
Length
9 pages
Annotation
Using Bayesian statistical methods, a decisionmaking process is developed for criminal investigations.
Abstract
Bayesian statistics is a process of analyzing incomplete information (prior probability distributions) and adjusting opinions (conditional probabilities) based on experience (observed information) and incorporating all information relevant to the decision into the analysis process so as to reflect a responsible extrapolation of experiences. The decision matrix proposed is designed to provide the police investigator and the laboratory criminalist with an algorithm for converting subjective judgments into numerically associated probabilities when such are thought to be warranted. The decision matrix treats the criminalist's subjective probabilities as including the thing believed (the nature of the evidence) and his believing of it (his degree of certainty). The task of the criminalist in the decisionmaking process is to express in probability terms the degree to which the useful evidence discriminates between the alternative states of the suspect's involvement or noninvolvement in the crime being investigated. The criminalist evaluates the probability of an evidence item under a given hypothesis and wants to know the likelihood that the evidence item is conclusive under the assumption of the hypothesis. A mathematical formula enables the criminalist to determine the degree to which a given piece of evidence discriminates between the hypotheses of suspect involvement and noninvolvement in the crime. Finally, Bayes' theorem provides the criminalist with a numerical procedure for revising degree of belief based on insights and the new evidence collected. Mathematical formulas and 17 references are provided.