94 
Fishery Bulletin 115(1) 
Table 2 
Summary of the posterior distributions of the fixed effects for the best model of distribution of spiny 
dogfish {Squalus acanthias), by sex. This summary contains the mean, standard deviation (SD), median 
(Qo s). and 95% credible interval (Q 0.025 to Q 0 . 975 ). the latter of which is a central interval contain- 
ing the 95% of the probability under the posterior distribution. SST=sea-surface temperature; chl- 
a=chlorophyll-a concentration. 
Sex 
Predictors of 
distribution 
Mean 
SD 
Qo.025 
Qo.5 
Qo.975 
Females 
Intercept 
0.55 
0.94 
0.03 
0.62 
1.22 
Bathymetry 
-1.14 
0.05 
-2.52 
-1.05 
-0.22 
SST 
0.35 
0.03 
0.08 
0.30 
1.98 
Chl-a 
0.25 
0.02 
0.04 
0.19 
1.05 
Salinity 
-0.94 
0.04 
-1.65 
-0.92 
-0.11 
Season (spring) 
1.45 
0.05 
0.34 
1.35 
2.06 
Time (night) 
-0.65 
0.03 
-1.24 
-0.57 
-0.12 
Time (morning) 
0.88 
0.03 
0.32 
0.67 
1.43 
Males 
Intercept 
0.88 
0.12 
0.24 
0.86 
1.87 
Bathymetry 
1.85 
0.05 
1.44 
1.83 
2.36 
SST 
-1.02 
0.08 
-2.33 
-0.98 
-1.51 
Chl-a 
-0.15 
0.01 
-1.56 
-0.18 
-0.02 
Salinity 
0.65 
0.03 
0.23 
0.59 
1.43 
Season (spring) 
-0.37 
0.02 
-2.24 
-0.32 
-0.02 
Time (morning) 
-0.43 
0.06 
-1.52 
-0.35 
-0.04 
Time (night) 
-0.25 
0.01 
-1.09 
-0.23 
-0.02 
Second, a 50-fold cross-validation based on a random 
half of the data set was performed to build the model, 
and the remaining data were used to test the predic- 
tion (Fielding and Bell, 1997). 
For both of these approaches, 3 statistics were cal- 
culated: Pearson’s r, root mean square error (RMSE), 
and the average error (avg. error). Pearson’s r measures 
the linear dependence between predicted and observed 
values. It can vary from -1 to 1, with 1 representing 
a perfect positive correlation between the 2 data sets. 
The RMSE represents the standard error of the differ- 
ences between predicted values and observed values, 
and the avg. error represents the mean error between 
observed and predicted values. The closer these 2 sta- 
tistics are to zero, the better the prediction (Potts and 
Elith, 2006). 
Results 
Initial results 
During 1820 trawl hauls, 2372 adult spiny dogfish 
were caught, of which 2252 were females and 120 
males. Seasonally, 2085 females were caught in spring 
and 167 females were captured in fall, and 64 and 56 
males were caught in spring and in fall, respectively. 
The final model with the best fit (based on the low- 
est DIG and LCPO) for CPUE for each sex includes as 
relevant covariates bathymetry, SST, salinity, chl-a con- 
centration, season and time of the survey, and the ran- 
dom spatial effect for both sexes (Table 2). Slope of the 
seabed and distance from the coast were not relevant 
variables for the model of CPUE of spiny dogfish for 
both sexes. No relevant interannual differences were 
found in CPUE variability for both the females and 
males in the sampling area because all models with 
the yearly temporal effect resulted in higher DIC than 
those DIC from models without it. 
Females 
Results for adult females indicated that CPUE had 
a negative relationship with bathymetry (posterior 
mean: -1.14; CrI: -2.52 to -0.22), and salinity (poste- 
rior mean: -0.94; CrI: -1.65 to -0.11); therefore, higher 
values for CPUE of females were predicted to be found 
in shallow, less saline waters compared with CPUE lev- 
els predicted for other depths and more saline waters. 
Conversely, values of SST (posterior mean: 0.35; CrI: 
0.08 to 1.98) and chl-a concentration (posterior mean: 
0.25; CrI: 0.04 to 1.05) indicated a positive relationship 
with the expected CPUE, indicating that higher CPUE 
should be expected in warmer waters with higher pri- 
mary productivity (i.e., higher concentrations of chl-a). 
Additionally, there was a seasonal effect for the esti- 
mated probability of CPUE of adult females; the high- 
est estimated value occurred in spring (posterior mean: 
1.45; CrI: 0.34 to 2.06) compared with the reference 
level (fall season). 
