Modeling uncertainty in the measurement of low-level analytes in environmental analysis

The use of analytical chemistry measurements in environmental monitoring is dependent on an assessment of measurement error. Models for variation in measurements are needed to quantify uncertainty in measurements, set limits of detection, and preprocess data for more sophisticated analysis in prediction, classification, and clustering. This article explains how a two-component error model can be used to accomplish all of these objectives. In addition, we present applications to quantitating biomarkers of exposure to toxic substances using gene expression microarrays.