Use of blood levels to infer carcass levels of contaminants

Inferences may be made about the carcass levels of a contaminant based on the contaminant level in blood samples. A method is given for comparing such populations that utilizes bivariate normal distributions and their principal axes, thereby avoiding a dilemma arising from the use of regression techniques. Confidence intervals and power calculations are given. Data from captive barn owls provide partial justification for the use of this method.