NCJ Number
213321
Journal
Canadian Society of Forensic Science Journal Volume: 38 Issue: 4 Dated: December 2005 Pages: 205-212
Date Published
December 2005
Length
8 pages
Annotation
Published models that define the ethanol partition coefficient as a function of temperature are widely used in the calibration and testing of breath-test instruments used to determine the amount of alcohol in a person's system; this study investigated further the statistical development of these published models.
Abstract
The original model derived from the data of Harger et al. apparently was developed inappropriately from ordinary least squares analysis of linear transformed data. Still, data that resulted from the work of Harger et al. is reliable and relevant for continued use in instrument testing and calibration; it just requires proper analysis. A model developed appropriately from nonlinear regression was not only in better agreement with other published models, but also resulted in the lowest variation in the instrument measurements. The maximum variation in measurements taken with the different models was 2.5 percent. Moreover, there were uncertainties associated with all of the model estimates. Despite these uncertainties, however, all of the models yielded coefficient estimates that were forensically acceptable. All of the models produced calibrated instrument results well within accepted standards for accuracy set by the U.S. National Highway Traffic Safety Administration and the Canadian Society of Forensic Science Alcohol Test Committee. 3 figures and 6 references