Abstract

When ages of fish are estimated via examination of their hardparts (i.e., otoliths, scales, rays, etc.), the precision (i.e., reproducibility) of those age estimates can be measured in several ways. Percent precision in our study represents the percent of replicated age estimates (i.e., for the same fish) that agree exactly or within some appropriately narrow age range (e.g., + 1 year). The average percent error and coefficient of variation are slightly different formulas designed to express the uncertainty of the average estimated age. One (two, or even all three) of these measures can be calculated for many fish to develop an index (or indices) of precision, which is used to evaluate the consistency for which ages have been estimated within a sample of fish. The correlation coefficient can also be used to measure the association between replicate ages for a sample of fish. All four measures may be used to evaluate use of a particular hard part or preparation technique to age fish or to compare the ‘ease’ of ageing one species versus another. The use of these measures is problematic for a variety of reasons, as has been shown for a number of “real” data sets from fishery labs. We used a simulated data set with different levels of variance to evaluate the utility of these four measures of precision. Using simulated data with known patterns of precision and bias that represent a number of anticipated scenarios has been missing in the discussion of the relative efficacy of these different measures of precision.

PDF Preview