LAI ET AL.: AGE DETERMINATION OF PACIFIC COD 



twice by the senior author, who had 3 years expe- 

 rience in age determination using all of the age- 

 ing methods. 



A loglinear model (Fienberg 1981) was used to 

 test the independence of repeatability (R ,k = 1,2 

 corresponding to agreement and disagreement) of 

 age readings corresponding to age-class (A , 

 J = 1,...,8 corresponding to age classes given 

 in Table 1) and ageing method (M, i = 1,...,5 

 corresponding to ageing methods in the sequence 

 given in Table 1). See Appendix A for the details 



Table 1 . — Observed frequency of agreement between two age 

 readings using different ageing methods. Percent agreement is 

 shown in parentheses. 



of this statistical method. The computer program 

 P4F in BMDP (Dixon 1983) was used for compu- 

 tation and analysis. 



An analysis of variance (ANOVA) model with 

 repeated measures (Winer 1971) was used to ex- 

 amine variability in age readings due to the 

 methods. The ANOVA model was 



Xy„ = M- + TT^j + M, + (M7T)j„ + Rj + (i?1T)j„ 



+ {MRTT)ijn +ey„ 



where, Xy„ is the observed age of the nth fish by 

 the ith ageing method and jth 

 reading, 



i = 1, 2, 3, 4, 5 indicates ageing 

 methods in the sequence shown 

 in Table 1, 



J = 1, 2 indicates the first and sec- 

 ond age readings, 



n = 1,...JSI indicates number of 

 fish, 



|JL is the grand mean, 

 ■n is the effect between subjects (indi- 

 vidual fish), 

 M is the effect of ageing method, 

 R is the effect of reading, 

 (Mir), {R tt), and (MR -it) are the two- and 

 three-factor interactions of effects M , 

 R , and tt, and 

 e is the random error. 



Readings from coracoids were not included in the 

 analysis because of the small sample size. Speci- 

 mens with missing values were also excluded 

 from this analysis. The Q -statistic (Snedecor and 

 Cochran 1967) was used to test the differences 

 between the mean ages between readings and be- 

 tween ageing methods. 



Validation of age determination was carried 

 out by comparing the mean lengths at age esti- 

 mated from the five ageing methods to the modal 

 lengths of the 1977 year class. The progression of 

 this year class could be traced by examining the 

 length-frequency distributions for 1978 to 1983. 



Because existing age data files for Pacific cod at 

 the Northwest and Alaska Fisheries Center 

 (NWAFC) were derived from scales, it is impor- 

 tant to explore whether or not the scale age data 

 can be used for stock assessment. A simulation 

 study was carried out to convert a scale age- 

 length key to a dorsal fin-ray age-length key. 



715 



