FISHERY BULLETIN; VOL. 83, NO. 2 



mension (width) to the nearest 0.05 mm. When the 

 left otolith was chipped or broken, the right one 

 was substituted for measurements, since no sys- 

 tematic differences between left and right otolith 

 measurements were apparent for either species. 

 The left otolith was subsequently sectioned and 

 age determined by the same otolith reader. Otolith 

 thickness, which is too variable to measure on the 

 whole otolith, was measured on the section from 

 internal to external surface just dorsal to the sul- 

 cus (Fig. 1). 



Confirmation Subsample 



In order to test the precision of the model, sub- 

 samples of 50 otoliths by sex and species were 

 drawn randomly from samples not used in the 

 calibration subsample. These samples were han- 

 dled in the following way: A second whole otolith 

 age was determined by reader A to determine 

 within-reader variability for S. diploproa and 

 between-reader variability for S. pinniger (reader 

 B had left this laboratory). The otoliths were sent 

 to the Northwest and Alaska Fisheries Center 

 (Seattle, Wash. ) for an additional whole otolith age 

 to determine between-agency variability. The 

 otolith was dried, weighed, measured, and sec- 

 tioned as described above; a single otolith section 

 age for each specimen was determined by reader A 

 for both species. Model-estimated ages were de- 

 termined by use of the multiple regression models 

 described below. 



Data Analysis 



Generally, data were recorded in a standard 

 format and stored on the Oregon State University 

 Cyber 70 computer. Data management and analy- 

 sis were assisted by use of the Statistical Package 

 for the Social Sciences (SPSS) (Nie et al. 1975). 



From the calibration subsample of otoliths, pre- 

 dictive regression equations were developed to es- 

 timate age from otolith morphometries. Multiple 

 regression models were fitted in the following 

 form: 



Age = 6iXi -I- 62^2 + 63X3 -I- b^X^ + c 



where age (years) is determined by conventional 

 methods, bn's = regression coefficients, X 's = in- 

 dependent variables, and c = constant. Models 

 were developed for males and females separately 

 within each species with both otolith section ages 

 and whole otolith ages as dependent variables. 



Independent variables included otolith weight, 

 otolith length, otolith width, the respective square 

 and cubic terms of each, and the interaction vari- 

 ables (otolith weight/otolith length and otolith 

 length/otolith width). With the exception of 

 otolith weight, where both weight and the cube of 

 weight were used as independent variables, 

 square or cubic terms were not used if the raw 

 values were entered. This decreased problems of 

 multicollinearity. Models were fitted in a forward 

 stepwise manner (Nie et al. 1975) with the inclu- 

 sion level for independent variables set at 

 P = 0.10. 



The 1980 confirmation subsample was used to 

 verify the models. Direct comparisons between 

 ages determined for the same otoliths but dif- 

 ferent reading methods were accomplished by 

 paired ^-tests. Since age is not known with cer- 

 tainty for any otolith, the ages determined by 

 reader A for S. diploproa and by reader B for S. 

 pinniger, which were used to calibrate the models 

 in the calibration subsample, were considered as 

 "standard age". To conduct multiple comparisons 

 of variability, deviations from standard age were 

 defined as follows: "model-induced variation" is 

 the difference between the standard age and the 

 model-predicted age; "within-agency variation" is 

 the difference between ages determined by reader 

 A for S. diploproa and between readers A and B for 

 S. pinniger; "between-agency variation" is the dif- 

 ference between the standard age and the age de- 

 termined by the National Marine Fisheries Ser- 

 vice (NMFS). A one-way analysis of variance 

 (ANOVA) was used to compare these deviations. 

 Multiple range testing was conducted using the 

 least significance difference method with 

 a = 0.05. This analysis was conducted only for 

 whole otoliths since only a single section age was 

 determined on the 1980 confirmation subsample. 



RESULTS 



Sehastes diploproa 



Locations of the collections of S. diploproa are 

 shown in Figure 2; this species was taken from lat. 

 36°49' to 48°47'N and over a depth range of 62 to 

 338 m. The distribution was similar to that noted 

 in 1977 (Boehlert 1980). A total of 975 male and 

 1,145 female specimens were taken during the 

 survey. The length frequencies show a mode near 

 23 cm for males and 24 cm for females with sec- 

 ondary modes at 26 and 27 cm, respectively. Cor- 

 responding age frequencies (based upon whole 



106 



