240 
Fishery Bulletin 117(3) 
Table 4 
Results for parameters of the von Bertalanffy growth func¬ 
tion for blackspotted rockfish ( Sebastes melanostictus) and 
rougheye rockfish (S. aleutianus) collected in the Gulf of 
Alaska in 2009 and 2013. The parameters are the mean 
asymptotic length (L„), the growth coefficient (K) and the 
time or age when mean length was zero ( t 0 ). n=number of 
specimens. 
Parameter 
Blackspotted 
rockfish 
Rougheye 
rockfish 
n 
546 
313 
L„ 
526 mm 
544 mm 
K 
0.062 
0.102 
to 
0.2 
-0.36 
that are correctly identified in the field. However, the 
model substantially improves identification of rougheye 
rockfish, for which field identifications are only 66-68% 
accurate. The results of a more recent study of field 
identifications from the season in 2015 indicate 75% 
accuracy (Shotwell et ah, 2017), but this level of accu¬ 
racy is still significantly worse than the results obtained 
in this study. 
Most studies of this kind will assign a species to all 
specimens, even if the model indicates a high degree 
of uncertainty. We recommend a more conservative 
Table 5 
Regression coefficients for the final form of the logistic 
regression model following stepwise elimination of param¬ 
eters. Data used in the regression model are from black- 
spotted rockfish ( Sebastes melanostictus) and rougheye 
rockfish (S. aleutianus ) collected in the Gulf of Alaska in 
2009. After stepwise elimination, all remaining parame¬ 
ters were significantly different from zero. 
Parameter 
Coefficient 
Error 
^-score 
P 
Intercept 
-18.92 
5.37 
-3.53 
<0.01 
Area 
1.07 
0.25 
4.29 
<0.01 
Perimeter 
2.88 
0.64 
4.51 
<0.01 
Minor axis length 
7.36 
3.50 
2.10 
0.04 
Otolith width 
-7.96 
3.12 
-2.55 
<0.01 
Otolith length 
-3.23 
1.06 
-3.06 
<0.01 
Otolith weight 
-174.87 
59.03 
-2.96 
<0.01 
Fork length 
-0.08 
0.02 
-4.60 
<0.01 
Log 10 (A£e) 
32.74 
10.82 
3.03 
<0.01 
Perimeter x 
Log 10 (Age) 
-2.81 
0.55 
-5.15 
<0.01 
Otolith weight x 
Log 10 (Age) 
133.79 
33.78 
3.96 
<0.01 
approach as a further defense against errors imposed 
by the process of otolith age determination. Our method 
trades a decrease in misclassifications (improved spec¬ 
ificity) for the cost of fewer correct classifications 
Table 6 
Results of species classification from the logistic regression model comparing the agreement between the 
genetic identification and the species predicted by the model constructed and validated with data for black- 
spotted rockfish (Sebastes melanostictus) and rougheye rockfish ( S. aleutianus ) collected in the Gulf of Alaska 
in 2009 and 2013. Data are from the training data set used to fit the regression model and from the testing data 
set created when the regression model was used on new data. Values in the left columns provide the results 
when all specimens were classified and those in the right columns indicate the effect of removing specimens 
classified as uncertain. n=number of specimens; ID=identification. 
Number of specimens (successful ID [%]) 
Model ID (all classified) Model ID (uncertain cases removed) 
Data set for 
Blackspotted 
Rougheye 
Blackspotted 
Rougheye 
genetic ID 
rockfish 
rockfish 
rockfish 
rockfish Uncertain 
Training data set 
Blackspotted rockfish 
424 
10 
411 
5 
18 
77=434 
(97.6%) 
(2.3%) 
(94.7%) 
(1.2%) 
(4.1%) 
Rougheye rockfish 
11 
193 
7 
180 
17 
re=204 
(5.4%) 
(94.6%) 
(3.4%) 
(88.2%) 
(8.3%) 
Testing data set 
Blackspotted rockfish 
110 
2 
109 
1 
2 
77 = 112 
(98.2%) 
(1.8%) 
(97.3%) 
(0.9%) 
(1.8%) 
Rougheye rockfish 
10 
99 
6 
94 
9 
71=109 
(9.2%) 
(90.8%) 
(5.5%) 
(86.2%) 
(8.2%) 
