Nonparametric estimates of age misclassification from paired readings

With a large enough sample, age misclassification probabilities can be estimated nonparametrically by fitting the observed distribution of differences between paired readings. Application of this procedure to Pacific halibut data reveals that the unsigned deviations of single readings from the mode follow a geometric distribution, leading to a simple specification of misclassification probabilities. The potential for bias in this and other published procedures is discussed.