554 



Fishery Bulletin 91(3), 1993 



larger sample size available from 

 the commercial fishery for fish in 

 these length classes. Based on the 

 lack of pattern among the signifi- 

 cance tests and the lack of tests ex- 

 ceeding a*, it appears that survey 

 gear does not more consistently se- 

 lect for fish of a different age at a 

 given length than do commercial 

 fishing gears. Accordingly data on 

 age at length obtained from fish col- 

 lected in the research surveys can 

 be combined with data from fish 

 sampled commercially. 



Catch at age and precision 



Catch-at-age estimates obtained 

 with the various age-length keys 

 showed no indication of systematic 

 bias (Fig. 2). This was expected, as 

 the results of statistical tests indi- 

 cated no significant difference be- 

 tween the age-length keys that were 

 pooled. The precision of catch-at-age 

 estimates, however, did show a 

 trend among the different levels of 

 aggregation (Fig. 3). Combining first 

 and second quarter age-length keys 

 derived solely from commercially 

 caught haddock increased precision 

 for all age groups, particularly for 

 older age classes that typically had 

 small sample sizes within a single 

 quarter (e.g., the 82 and 84cm 

 length classes; Table 1). The inclu- 

 sion of survey age data had a rela- 

 tively smaller effect on the precision 

 of estimates for most age classes ex- 

 cept for age-2 fish (Fig. 3). This oc- 

 curred because the sample size of 

 small haddock was generally less 

 in the commercial samples than in 

 the research survey samples. 



Statistical considerations 



Fisher's exact test tests the hypoth- 

 esis that the proportion at age 

 within each length class is no dif- 

 ferent among keys than would be 

 obtained by random chance. 



Formally, the hypothesis tested 



