182 
Fishery Bulletin 117(3) 
Table 1 
Multiple brooder classification (MB; Y=yes, N=no), mean latitude (in decimal degrees), maximum fork length 
(L„, cm), observed maximum age (A^^, years), and natural mortality rate (M) for each species of rockfish 
(Sebastes spp.) collected by the Pacific Fishery Management Council and NOAA Southwest Fishery Science 
Center between 2004 and 2015 on the continental shelf along the West Coast of the United States. For species 
for which parameters varied by sex, values for females are given. 
Common name 
Scientific name 
MB 
Mean latitude 
Amax 
M 
Bocaccio 
Sebastes paucispinis 
Y 
35.87 
75.90 
37 
0.15 
Canary 
Sebastes pinniger 
N 
44.78 
62.00 
84 
0.05 
Chilipepper 
Sebastes goodei 
Y 
37.82 
52.00 
35 
0.16 
Cowcod 
Sebastes levis 
Y 
35.50 
87.00 
55 
0.06 
Flag 
Sebastes rubrivinctus 
N 
34.24 
51.00 
38 
0.12 
Greenblotched 
Sebastes rosenblatti 
Y 
34.06 
57.99 
50 
0.09 
Greenspotted 
Sebastes chlorostictus 
Y 
36.53 
44.20 
51 
0.09 
Greenstriped 
Sebastes elongatus 
Y 
42.86 
37.26 
54 
0.08 
Halfbanded 
Sebastes semicinctus 
N 
34.57 
18.14 
15 
0.26 
Pink 
Sebastes eos 
Y 
34.00 
56.00 
66 
0.07 
Pygmy 
Sebastes wilsoni 
N 
42.67 
23.00 
26 
0.26 
Redstripe 
Sebastes proriger 
N 
45.53 
61.00 
55 
0.08 
Rosethorn 
Sebastes helvomaculatus 
N 
43.78 
28.66 
87 
0.05 
Rosy 
Sebastes rosaceus 
Y 
33.87 
32.90 
18 
0.27 
Shortbelly 
Sebastes jordani 
Y 
36.00 
28.50 
32 
0.26 
Silvergray 
Sebastes brevispinis 
N 
45.78 
71.00 
82 
0.05 
Speckled 
Sebastes avails 
Y 
33.67 
49.99 
37 
0.13 
Squarespot 
Sebastes hopkinsi 
Y 
34.82 
25.25 
19 
0.26 
Starry 
Sebastes constellatus 
Y 
35.64 
45.00 
32 
0.15 
Stripetail 
Sebastes saxicola 
N 
38.30 
33.05 
38 
0.12 
Swordspine 
Sebastes ensifer 
Y 
33.76 
17.60 
43 
0.11 
Vermilion 
Sebastes miniatus 
N 
34.73 
62.40 
60 
0.07 
Widow 
Sebastes entomelas 
N 
42.90 
50.34 
60 
0.08 
Yellowtail 
Sebastes flavidus 
N 
46.37 
52.20 
64 
0.15 
of both data sources, for each species. We then pooled 
data from the respective data sets over the study period 
(2004-2015) and summarized them by calculating the 
arithmetic mean to provide estimates of mean latitude 
(in decimal degrees), mean temperature at depth (in 
degrees Celsius), mean DO concentration at depth (in 
milliliters per liter), mean survey depth (in meters) for 
each species in the study (n=l, for each species). We fit 
binomial generalized linear models to the probability of 
each species being identified as a multiple brooder. The 
capacity to produce multiple broods was the response 
variable, and the environmental and demographic met¬ 
rics were predictor variables. 
Because temperature and latitude were likely highly cor¬ 
related, we analyzed the collinearity between all explana¬ 
tory variables using scatterplots and conducted a variance 
inflation factor (VIF) analysis. In the VIF analysis, we 
used an a priori cutoff of 10 (Craney and Surles, 2002) 
to assess whether collinearity between the variables was 
problematic. If variables had a VIF greater than 10, they 
were not used in the same model. Using the results of the 
VIF analysis, we created 23 candidate models that repre¬ 
sented hypothesized potential predictor-response relation¬ 
ships. We compared models using an information-theoretic 
approach using Akaike information criterion corrected for 
small sample sizes (AIC c ) (Burnham and Anderson, 2002). 
Following the general method used by Kowalski et al. 
(2015), we computed the maximized log-likelihood, AIC 
and AIC c scores, AAIC c values (difference between the AIC,. 
of model i and the smallest AIC c among the considered 
models), and the AIC weight (AIC W ) for each model. A pri¬ 
ori, we decided to describe the level of empirical support for 
a model using the terms substantial, considerably less, and 
essentially none to correspond to AAIC c values less than 2, 
between 4-7, and greater than 10, respectively (as advised 
by Burnham and Anderson, 2002). 
We evaluated the relative importance of influence of 
each predictor by summing the product of AIC W and fre¬ 
quency of occurrence of the variable for all models. We 
computed the log evidence ratio (LER) in support for each 
predictor by calculating the log 10 of its evidence ratio (quo¬ 
tient of relative importance and its complement) and used 
LERs as the basis of inference for predictor comparisons. 
Following Kass and Raftery (1995), a priori, we decided 
to use the terms minimal, substantial, strong, and deci¬ 
sive to correspond approximately to LERs greater than 
0, 0.5, 1, and 2, respectively. All analyses were executed 
by using R statistical software (vers. 3.4.4; R Core Team, 
2018), including the glm, AIC, and ggplot functions of this 
software. 
