212 
Fishery Bulletin 1 14(2) 
Table 3 
Insample generalized least squares (GLS) models and the integrated GLS and vector autoregression 
error correction models for the growth of age0.3 male chum salmon (Oncorhynchus keta) from Quil- 
cene River, Washington. Growth indices included 1 st immature year, 2 nd immature year (SW3), and 
maturing (SW4) life stages. SST=sea surface temperature. PDO=the Pacific decadal oscillation index. 
lAs,chum,t=i mma t ure Asian chum salmon. Iec-SE.t immature chum and pink salmon abundance from 
British Columbia to southeast Alaska. M A sch U m,t =ma t ur ing chum salmon abundances from Asia. Model 
statistics included coefficient estimates (coeff. ), P-values of the coefficients (P-value), the coefficient of 
determination (r 2 ), P-value of the P-statistic, coefficient of variation (CV), and Schwarz information 
criterion (SIC). 
Variables 
Model statistics 
Response 
Predictors 
Coeff. 
P-value 
r 2 
P-value 
CV 
SIC 
SW2 
I AS, Chum, t 
-0.562 
<0.001 
0.35 
<0.001 
0.054 
214 
SW3 
f-BC-SE.t 
-0.797 
<0.001 
0.52 
<0.001 
0.070 
196 
SW3 
f-BC-SE.t 
-0.835 
<0.001 
0.73 
<0.001 
0.052 
183 
SST t _ 2 
0.450 
<0.001 
PDO t _ 2 
-0.434 
<0.001 
SW4 
M AS, Chum, t 
0.718 
<0.001 
0.52 
<0.0001 
0.063 
202 
rioration (F-det.). The growth model for maturing fish 
did not deteriorate, but the coefficient of determination 
was close to zero. 
According to the results of the Lehmann’s correla- 
tion test (Table 6), the mean squared errors were sig- 
nificantly reduced for both insample regression equa- 
tions when integrating the error correction equation. 
However, the mean squared error was not significantly 
reduced for any model when applied to the reserved 
observations. 
From the SECM survey, at the station with the 
highest annual catch of juvenile pink salmon and chum 
salmon (>250 fish), we found an inverse relationship 
between the interannual mean length of chum salmon 
and catch of juvenile pink and juvenile chum salmon. 
Discussion 
Our study advances the understanding of potential fac- 
tors influencing marine growth of chum salmon in the 
North Pacific Ocean. With this study, we have contrib- 
uted two more chum salmon populations, 1 from the 
southeast Alaska and 1 from Washington state, to an 
Table 4 
Validation statistics based on the application of the growth models for age-0.3 male chum salmon ( On- 
corhynchus keta ) from Fish Creek, Alaska to the reserved observations. GLS=the generalized least 
squares regression model. Growth indices included middle juvenile (SWlb), late juvenile (SWlc), 1 st 
immature year, 2 nd immature year (SW3), and maturing (SW4) life stages. GLSWAR was the GLS and 
vector autoregression-integrated model. Statistics include sample size ( n ), coefficient of variation (CV), 
coefficient of determination (r 2 ), P-statistic, P-value of the P-statistic, P-statistic of model deterioration 
(F-det.), P-value of P-det. (P-det.), and the Schwarz’s information criterion (SIC). Negative r 2 values and 
associated P and P values were not shown for the SWlc, SW2, SW3 and SW4 GLS models. 
SWlb 
SWlc 
SW2 
SW2 
SW3 
SW4 
SW4 
GLS 
GLS 
GLS 
GLSWAR 
GLS 
GLS 
GLSWAR 
n 
6 
7 
7 
7 
7 
7 
7 
CV 
0.05 
0.14 
0.07 
0.07 
0.12 
0.05 
0.04 
r 2 
0.25 
0.27 
P 
1.01 
1.32 
P 
0.44 
0.36 
P-det. 
1.47 
1.41 
2.38 
6.10 
3.64 
0.77 
0.69 
P-det. 
0.23 
0.25 
0.05 
<0.0001 
0.008 
0.61 
0.68 
SIC 
34 
51 
56 
63 
58 
39 
41 
