NOTE Hayes A statistical method for evaluating age-length keys for Melanogrammus aeglefinus 



553 



LLI 60 



JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC 



MONTH 



Figure 1 



Monthly mean length at age of Georges Bank haddock 

 Melanogrammus aeglefinus in the commercial catch. 1980- 

 1988. 



Significant differences were observed between sec 

 ond and third quarter age-length keys; 45 of 95 com 

 parisons exceeded the 0.05 level, and 19 

 tests of these tests had probabilities of 

 less than 0.00054 (a) (Table 3). Tests be- 

 tween the third and fourth were not as 

 definitive; 21 of 85 comparisons exceeded 

 the 0.05 level, but no tests exceeded 

 0.00060 (a). Several tests, however, were 

 close to a' (Table 4). The large number of 

 tests exceeding 0.05 and the occurrence 

 of contiguous blocks of tests with low 

 probability (Table 4) indicate that age- 

 length keys differ between the third and 

 fourth quarters, but the power to detect 

 these differences may have been limited 

 by small sample sizes. In general, the 

 number of fish aged from the fourth quar- 

 ter was lower than those from other quar- 

 ters as indicated by the smaller number 

 of length intervals with sample sizes 

 greater than six. From these results, age- 

 length keys should not be combined 

 across the second and third quarters or 

 third and fourth quarters for haddock. 

 The consequence of pooling age-length 

 keys across quarters when statistical 

 tests indicate significant differences 

 would be to bias catch-at-age estimates 

 (Westrheim and Ricker, 1978). In the case 

 of Georges Bank haddock, the effect of 

 pooling third quarter age-length keys 

 with second quarter keys (for example) 



would be to bias catch-at-age estimates for the second 

 quarter towards younger fish. This occurs because 

 growth during the third quarter tends to increase the 

 mean length at age, and conversely, fish of a given 

 length tend to be younger during the third quarter 

 than during the second quarter. 



Comparison of pooled first and second quarter age- 

 length keys with spring NEFSC bottom trawl surveys 

 keys resulted in 7 of 71 tests having probabilities less 

 than 0.05 (Table 5), but none exceeding the a level of 

 0.00072. Although more "significant" results were ob- 

 tained than would be expected by random chance, no 

 consistent pattern among these differences was appar- 

 ent (Table 5). Further, several of these differences oc- 

 curred in the 60 to 66 cm size classes where the sample 

 size from the bottom trawl survey was generally small 

 ( < 10 fish ), while sample sizes from the commercial catch 

 were often large (>50 fish). Examination of the contin- 

 gency tables for these size classes indicated that the 

 difference in proportion at age between commercial 

 and survey age-length keys was small and was often 

 due to a broader representation of age classes in the 

 commercial data. A broader representation in the com- 

 mercial samples would be expected, however, given the 



