340 
Fishery Bulletin 115(3) 
cost in estimating sample sizes required for age com- 
position (Schweigert and Sibert, 1983; Lai* 1987), how- 
ever, the data on cost for the species sampled by the 
AFSC was not available at the time of this study. 
Intrahaul correlation arises owing to the similarity 
of fish ages within a given haul or the spatial distribu- 
tion of these ages in comparison with the spatial dis- 
tribution of sampling, which then leads to over-disper- 
sion of uncertainty when compared with what would 
be determined from multinomial sampling (McAllister 
and lanelli, 1997). Intrahaul correlation in sampling 
for ages has recently been the subject of several stud- 
ies (Pennington et ah, 2002; Hulson et ah, 2011), as 
well as investigations of how to account for intrahaul 
correlations in SCAA models (Francis, 2011; Maunder, 
2011; Hulson et ah, 2012). Although intrahaul correla- 
tion has received attention in the literature in terms of 
integration into stock assessment models, it is not clear 
at this point in time how intrahaul correlation could 
be incorporated in estimating optimal distribution of 
age-composition sample sizes across species before a 
stock assessment. The magnitude of the difference be- 
tween effective sample size and the actual sample size 
collected is influenced by the age aggregations within 
schools or the spatial distribution of the species sam- 
pled, which may not be consistent across species. This 
should be a topic for future consideration but is beyond 
the scope of the current study. 
Value in this case would be defined as the value of 
the fishery, the sampling efforts of which are support- 
ing stock assessment. Using fisheries assessed by the 
AFSC as an example, the walleye pollock fishery in the 
eastern BS supports one of the largest and most valu- 
able groundfish fisheries in the world (lanelli et al. 7 ). 
The stock would require a decrease in age-composition 
sample size if the results of the current study were 
implemented in the AFSC bottom trawl age-sampling 
design. Including value into the sampling theory meth- 
od has not been previously explored but, mathemati- 
cally speaking, it could be implemented in a similar 
manner to that of cost. It is more challenging, however, 
to determine how age sample size affects the potential 
value of a fishery The results of the simulation analy- 
sis show that changing sample size in age composition 
affects the resulting uncertainty in an SCAA model. 
Methods have been proposed that would take into ac- 
count uncertainty when setting management quantities 
such that when uncertainty in SCAA model results in- 
crease, the harvest target rate decreases to account for 
this uncertainty (Prager and Shertzer, 2010). The east- 
ern BS pollock assessment employs a buffer based on 
the uncertainty of the estimation of the harvest target. 
Therefore, if sample sizes were decreased and SCAA 
model uncertainty increased substantially, the peter- 
7 lanelli, J. N., T. Honkalehto, S. Barbeaux, and S. Kotwicki. 
2015. Assessment of the walleye pollock stock in the East- 
ern Bering Sea. In Stock assessment and fishery evalua- 
tion report for the groundfish resources of the Bering Sea / 
Aleutian Islands regions. North Pacific Management Council, 
Anchorage, AK. [Available from website.] 
tial value of the fishery could decrease. A more rigorous 
analysis of value would have to include the numerous 
other factors that are a part of the overall value of 
a fishery (e.g., a decrease in quota could increase the 
price per kilogram or increase long-term value). Age- 
composition sample size may indeed be a very small 
factor in terms of value, but value is unquestionably 
one of the main factors influencing how age-composi- 
tion sample sizes are currently allocated by the AFSC. 
The simulation analyses with the SCAA model pro- 
vide guidance on the factors to consider when adjust- 
ing age-composition sample sizes in a multispecies 
data collection program. The results suggest that life- 
history and survey index uncertainty play key roles in 
determining the magnitude of influence that changing 
age-composition sample size has on SCAA model uncer- 
tainty The results of the simulation analysis indicated 
that age-composition sample size has a greater impact 
on the resulting uncertainty in SCAA models for spe- 
cies with high recruitment variability or low survey in- 
dex uncertainty, or both. In contrast, for species with 
low recruitment variability or high survey index uncer- 
tainty, or both, changes in age-composition sample size 
have a smaller influence on the resulting uncertainty 
from a SCAA model. Returning to the eastern BS wall- 
eye pollock example, the recruitment variability of this 
stock is higher than that of most species, and it has 
intermediate survey index uncertainty (intermediate 
between rockfish and flatfish species); therefore, de- 
creasing age-composition sample size would potentially 
have a larger impact than decreasing sample size for a 
flatfish species, for example, that has lower recruitment 
variability and low survey index uncertainty. 
To isolate the effect of the fishery-independent sur- 
vey data sources (index and age composition) we made 
several simplifying assumptions in our simulation. 
These involved including process and observation er- 
ror in the fishery data sources (e.g., different catch his- 
tories and different levels of uncertainty in the catch 
data) which would also influence the uncertainty re- 
sulting from an SCAA model and could decrease the 
relative influence of the fishery-independent survey 
data sources. An additional consideration that was 
not made in our simulation is the potential for gaps 
in the fishery-independent survey data (which is the 
case for the AI and GOA bottom trawl survey data) 
and how that influences age-composition sample size 
in the resulting SCAA model uncertainty The strength 
of the relationships between increasing or decreasing 
age-composition sample size and recruitment vari- 
ability and survey index uncertainty is possibly due 
to the simplifying assumptions made in the simula- 
tion analysis. Although, magnitudes in changes to the 
SCAA model uncertainty could be different with the 
use of more sophisticated simulations, we hypothesize 
that these correlations may be qualitatively the same 
regardless of the complexity of the simulation analysis. 
We recommend that future research into investigating 
the influence of age-composition sample size on SCAA 
model results, and how that relates to optimal distri- 
