28 
Fishery Bulletin 119(1) 
— Eberbach projector -- Microscope 
Estimate 
although there was some bias in the 
age estimates for Gulf menhaden and 
in estimates between readings made by 
using the Eberbach projector, the pre- 
cision of age estimates made with the 
microscope, in comparison with those 
made with the Eberbach projector, is 
sufficient to allow the menhaden pro- 
gram to update the technology used for 
aging these 2 menhaden species. Contin- 
ued aging for stock assessments should 
yield the same quality of data through- 
ee 
2 
Age 
— Eberbach projector -- Microscope 
g 
© 
£ 
a 
Lu 
out the time series. 
Sample sizes in our study were larger 
for each species than in many other 
studies that have considered aging pre- 
cision. Experts have suggested a sample 
size of 200 fish, 100 from a reference col- 
lection and 100 from recent production 
samples, is sufficient to identify signifi- 
cant differences when testing for aging 
errors or inconsistencies (Campana, 
2001). The large sample size in this 
study could have influenced the results 
of the bias tests and measures of aging 
2 
Age 
Figure 3 
Simultaneous multinomial confidence intervals of age estimates, by method of 
aging, for all samples of (A) Atlantic menhaden (Brevoortia tyrannus) and (B) 
Gulf menhaden (B. patronus) collected along the Atlantic and Gulf coasts of 
the United States. Samples included Atlantic menhaden collected in 2013 and 
2017, Gulf menhaden collected in 2005 and 2017, and reference collections for 
each species. Aging methods included the use of either an Eberbach projector 
or a stereo microscope. The solid line represents estimates of the proportion of ~ 
samples at each age determined with the projector, and the dashed line rep- 
resents estimates of the proportion of samples at each age determined with a 
microscope. Error bars indicate 95% confidence intervals. 
appropriate and applied in a timely manner. Quality 
assurance and QC help to determine whether perceived 
changes in age estimates are from variation in aging pre- 
cision or from real changes in fish populations (Morison 
et al., 1998). Although this study served as QA and QC 
for the Beaufort Laboratory's menhaden program, its 
main purpose was to assess the level of continuity in the 
data if age readings were done on a microscope rather 
than with the Eberbach projector. 
The menhaden program of the Beaufort Laboratory 
has been run without appreciable change for more than 
half a century, and well over a million samples have been 
amassed with a remarkable continuity in collection meth- 
ods and output. Because of the shift in aging methods that 
was necessitated by critical program updates, we needed 
to determine whether outputs from the use of the new 
method would be consistent with those from the use of the 
previous method. Results from our study indicate that, 
consistency. McBride (2015) used a 
sample size of 5 fish to produce 5 age 
estimates per age class, a sample size 
that like those in most studies is much 
smaller than the sample sizes used in 
our study. For example, other studies 
have used sample sizes of 85-180 per 
species when comparing aging struc- 
tures (Khan and Khan, 2009; Khan 
et al., 2015; Baudouin et al., 2016; 
Kumbar and Lad, 2016), 14-390 per 
species when comparing ages between 
readers (Wakefield et al., 2017; Khan 
et al., 2019), and 30-182 per species 
when comparing both aging methods 
and ages between readers (Giirsoy et al., 
2005; Goldman and Musick, 2006; Herbst and Marsden, 
2011). Given these smaller sample sizes, comparing age 
estimates from our study to those from the other studies 
has been difficult. 
How large should sample sizes be to determine if bias 
occurs? If we consider only data from the reference collec- 
tions of each species, the conclusions would be different 
than those from consideration of data from the full sample. 
For example, the PA is much higher for estimates based on 
the reference collections alone than for those based on the 
full sample, and the APE and ACV are much lower. The 
ages from the reference collections do not indicate signifi- 
cant bias for either species. Because of the smaller sample 
sizes, use of the reference collection for each species led to 
different conclusions than those from the use of the full 
sample; however, the sample sizes of the reference collec- 
tions are similar to the sample sizes that have been used 
in other aging studies. 
