Is Otolith Chemistry Useful for Examining Stock Mixing?

Otolith chemistry has been routinely used for stock discrimination and to infer movements for over 20 years. Technical and procedural methods have been well developed; however, statistical methods have only recently begun to be explored in sufficient detail to inform management. Rarely are differences among groups strong enough to permit near certainty in assignment probabilities. We used simulated data from two populations at three differences (low, medium, and high) to determine at what level of mixing the two populations would become indistinguishable. These populations were drawn from a multivariate normal distribution and homogenous variance-covariance matrix. Each scenario was simulated 10,000 times and the data were tested with a MANOVA and the classification accuracy was determined with a linear discriminant function.  Results from are important for evaluating studies where training data exhibit less than complete separation among populations, yet go on to apply their discriminant rule to classify unknown origin individuals. While real data are often less than perfect, knowledge of data limitations will aid in effectively evaluating competing data sources in stock discrimination studies.