Statistical analysis of blood- to breath-alcohol ratio data in the logarithm-transformed and non-transformed modes.
The statistical analysis of non-transformed and logarithm-transformed blood- to breath-alcohol ratios ("blood/breath ratios") is detailed. The data analyzed were derived from 137 simultaneous blood-alcohol and breath-alcohol concentration measurements made between 15 and 179 min after the end of drinking, with 136 of the measurements obtained during the 15- to 124-min time frame. Although the distribution of the non-transformed ratios is positively skewed, and that of the logarithm-transformed data more closely approximates the normal distribution upon visual inspection, both analyses generated results that do not differ significantly from each other when considered in the context of "mean ratios + or - 2SD". This is in accord with the results of the Kolmogorov-Smirnov goodness-of-fit test, which does not reject either dataset and demonstrates that both are approximately normal. Since the logarithm-transformed data generate more conservative statistical blood/breath ratio ranges than the non-transformed data, they were selected as the basis for the principal conclusion of this work. That conclusion is a refutation of the argument that, breath-alcohol analyzers relying on a 2100:1 blood/breath ratio tend to underestimate the blood-alcohol concentrations of driving-while-intoxicated arrestees because the commonly accepted mean postabsorptive ratio is 2300:1. In fact, whenever the absorption status of a driving-while-intoxicated arrestee at the time of a breath test cannot be definitively established, the results of this work support the application of a relative error range of -40% to +28% for 95% of the population, based on a statistical blood/breath ratio range of 1259:1 to 2679:1, and -46% to +42% for 99% of the population, based on a statistical range of 1128:1 to 2989:1.