NCJ Number
214532
Journal
Canadian Society of Forensic Science Volume: 39 Issue: 1 Dated: March 2006 Pages: 15-24
Date Published
March 2006
Length
10 pages
Annotation
This article proposes a decision rule to identify unacceptable differences in duplicate breath samples in drunk driving cases.
Abstract
In establishing a duplicate test agreement criteria or decision rule for breath alcohol analysis it is important to balance the probability of false rejection with the probability of error detection. The decision rules currently used do not adequately balance the probability of a false positive error with that of error detection nor do they account for the non-constant variance of breath alcohol measurement. The author suggests that minimizing the probability of false rejection can be accomplished by increasing the size of the critical difference decision rule, which in turn reduces the probability of error detection. This will decrease the risk of the more significant error, the false negative error, by increasing the ability to detect bias. The author presents the statistical basis for determining an appropriate decision rule and recommends a decision rule that accommodates the measurement variability through the use of normal distribution theory. Microsoft Excel is recommended as a good program in which to manipulate various decision rules in order to observe the effect of different rules on error probabilities. Equations are provided for these simulations. Research methods involved evaluating 4 decision rules by employing 3-digit duplicate breath test data (n=4,339) collected in Washington State over a 3-month period in 2005 using the BAC Datamaster. Equations, figures, table, references