Examples of Labs Accuracy and Precision Assessments
Last modified on Mar 05, 2019
MacLellan and Gillespie (2015) described how the Department of Fisheries and Oceans Canada (DFO) Schlerochronlology Lab uses verification and validation in age estimation studies. Validation refers to the accuracy of ages assigned, while verification refers to precision (Beamish and McFarlane 1983). DFO utilizes several methods to address these concepts. Validation is based on known-age, mark-recapture fish, whereas verification is achieved by using multiple readers and training (and testing) methods.
Most labs use some type of multiple independent reads to assess precision. For example, DFO has samples read independently by two readers (double blind). Readers assign ages without reference to fish length but use auxiliary information, such as location and collection date. If more than two readers are estimating ages of a location, samples are divided evenly among readers. Therefore, every reader ages a similar number of samples and will be paired with every other reader. Differences in age estimations between readers are resolved by a third reader viewing the sample with the two primary readers. They collectively resolve the differences and assign a final (consensus) age. If a final age cannot be determined, the sample is considered unreadable and removed from analysis. A comparison of length and age data complete the quality control process. Outliers are re-examined, and ages corrected, if an error occurred. Outliers may be discarded if determined to be data or sample collection errors.
Both Oregon Department of Fish and Wildlife (ODFW; Borgerson et. al. 2014) and Idaho Fish and Game (IDFG; Wright et al. 2014) use a similar process by having two trained readers conduct independent reads. Reads are blind with no knowledge of biological data. Similar to DFO, auxiliary data, such as basin, date, species, and stock information, are known. Unlike DFO, once reads are completed, the two readers review disagreements and confirm age. After ages are assigned, data are plotted to identify outliers, and body lengths are plotted against age estimates. Sometimes this causes an age to be changed, but most often, it results in notation of a potential data discrepancy. Once completed, ages are validated by checking for coded wire tags (CWTs). Although providing some validation, only total age is known from a CWT.
Beamish, R.J. and McFarlane, G.A. 1983. The forgotten requirement for age validation in fisheries biology. Transactions of the American Fisheries Society 112: 735-743.
Borgerson, L., Clemens, B., Bowden, K., and Gunckel, S. 2014. Fish life history analysis project: methods for scale analysis. Information Report 2014-10, Oregon Department of Fish and Wildlife, p. 41.
MacLellan, S. and Gillespie, D. 2015. Chinook Salmon (Oncorhynchus tshawytscha) Scale Age Determination Procedures. Canadian Technical Report of Fisheries and Aquatic Sciences 3123, 69.
Wright, K.K., Hernandez, K., Hohman, C., Reinhardt, L., Schrader, W., and Copeland, T. 2014. Laboratory methods for assigning ages to anadromous salmonid scale sample. Report to U.S. Department of Energy, B.P.A., Boise, ID.