FISHERY BULLETIN: VOL. 83, NO. 2 



Table 2. — Correlation matrix for selected otolith morphometric, weight, and age data 

 for the calibration subsample of Sebastes diploproa. 



length and the interaction variables compensate 

 for these changes in the pattern of otolith weight 

 in the multiple regression models offish age. 



The multiple regression models relating fish age 

 with otolith data were fitted with both whole 

 otolith age and otolith section age as dependent 

 variables. Independent variables included in the 

 whole otolith age models, their coefficients, and 

 significance levels are presented in Table 3. All 

 coefficients were highly significant and the models 

 explain 88.1% of the variation in age for females 

 and 92.0% for males, as measured by the coeffi- 

 cient of determination, R"^. Residuals from the 

 models by age category show no trend up to age 35 

 for females and age 30 for males, after which there 

 is a trend of increasing positive deviation with 

 increasing age. The ages included in this part of 

 the model, however, represented only 7.7% of 

 female and 8.6% of male S. diploproa and are 

 therefore not of great concern. These deviations 

 are positive, however, suggesting that the model 

 predictions may relate to otolith growth patterns 

 which are more indicative of otolith section ages. 



Variables included in the otolith section age 

 models, their coefficients, standard errors, and 

 significance levels are presented in Table 4. Again, 

 all coefficients are highly significant, but the co- 

 efficients of determination are slightly less, ex- 

 plaining 86.1% of the variation in age for females 

 and 85.0% for males. Mean residuals for the dif- 

 ferent age categories show no significant trend 

 with age. 



The model based upon whole otolith ages suffers 

 from inaccuracies in the older ages, where otolith 

 section ages are much greater than whole otolith 

 ages. This is demonstrated in the trend of increas- 

 ing residuals with increasing age. The model 

 based upon otolith section age, however, is charac- 



108 



Table 3. — Regression coefficients and associated statistics on 

 the multiple regression models of whole otolith age for Sebastes 

 diploproa . 



Table 4. — Regression coefficients and associated statis- 

 tics on the multiple regression models of otolith section 

 age for Sebastes diploproa . 



Variable 



Coefficient 



SE 



Females (A/ = 290) 

 Otolith weight 

 (Otolith width) 2 

 (Otolith weight)^ 

 (Otolith length)^ 

 Constant (a) 



SD = 4.232 

 Multiple correlation, R 



Males (W = 246) 

 Otolith weight 

 Otolith width^ 

 (Otolith weight)^ 

 (Otolith length)^ 

 Constant (a) 

 SD = 4.620 

 Multiple correlation, R 



0.2270 



-0.3288 



-0.1134 X 10-' 



-0.1114 



5.0243 



= 0.928 



0.2496 



-5.7233 



-0.1315 X 10-' 



-0.0882 



23.540 



0.922 



0.0137 

 0.0377 

 0.155 X 10- 

 0.0205 

 1 .2982 



00158 

 0.6949 

 0.266 X 10- 

 0.0256 

 3.3823 



<0.001 

 <0.001 

 <0.001 

 <0.001 

 <0.001 



0.001 

 0.001 

 0.001 

 0.001 

 <0.001 



terized by slightly lower multiple correlation co- 

 efficients (Table 4). This may be a result of inac- 

 curacies in estimates of otolith section age of 

 younger fish, where greater difficulty in age de- 



