area of collection would lead to time-area 

 variations of catches within all the sampled 

 trips to be the major source of error. 



We shall utilize these estimates of subsam- 

 ple variations to test the significance of 

 sample-to-sample variation in subsequent 

 analyses. 



Samples (between trips) 



Analyses of covariance among samples were 

 computed for each cell (each combination of 

 given year, area, month, and size category) 

 containing more than one sample (cf. Table 1). 

 The pooled analysis of covariance showed 

 significant adjusted mean differences among 

 samples, or trips, for both large and scrod size 

 categories (Appendix Table A3). The among 

 sample mean squares of large and scrod had- 

 dock for this pooled analysis (0.0364 and 

 0.0369) were greater than that among the five 

 samples used in the analysis of subsample 

 variation (0.0065, cf., Table A2). This may 

 have occurred because the five special samples 

 came from a more restricted time and area 

 within the sampling area than the general 

 samples. The among sample mean square is also 

 about five times larger than the within sample 

 or common mean squares which are used for 

 testing in a one-stage analysis. 



COMPARISON AMONG FACTOR LEVELS 



Size Categories 



To determine whether separate length- 

 weight equations should be used for scrod and 

 large haddock, covariance analyses were com- 

 puted for 16 trips from which both size 

 categories were sampled. The pooled analysis is 

 presented in Appendix Table A4; significant 

 differences were found for adjusted means. 

 Only subsample variation need be accounted 

 for in this analysis as comparison was between 

 large and scrod samples from the same boat. 



The adjusted means were calculated and 

 compared for each of these pairs of regression 

 equations. In all cases the adjusted mean was 

 greater for large than for scrod haddock (Table 

 2). The observed differences are to be expected 

 if the fish were sorted primarily on the basis of 

 heavy appearance, i.e., within the range of 

 cull-sizes the short, plump fish would be 

 considered large whereas the longer, slender 



Table 2— Natural logarithms of adjusted mean weights 

 (pounds) for samples of large and scrod haddock. 



Pair 



Adjusted means 



Adjusted means 



Number 



for large haddock 



for scrod haddock 



1 



0.8117 



0.7597 



2 



1.2468 



1.2221 



3 



0.8384 



0.8359 



4 



1.0587 



0.9788 



5 



0.7705 



0.7378 



6 



1.0844 



1.0240 



7 



0.9742 



0.9438 



8 



0.8334 



0.7952 



9 



1.0232 



0.9705 



10 



1.1383 



1.1261 



11 



1.1332 



1.1171 



12 



1.0552 



0.9996 



13 



1.1713 



0.9983 



14 



1.0661 



0.9674 



15 



0.6554 



0.6228 



16 



1.1104 



1.0369 



individuals would be classed as scrod. 



Years 



An analysis of covariance among years was 

 computed within each month, area, and size 

 category classification containing samples from 

 two or more years. For example, comparisons 

 between 1931 and 1932 were made for the 

 western Georges Bank area in each of the 

 months January, June, and July. A single 

 regression equation was used for each year, 

 combining several samples where required. The 

 several analyses were then pooled and no 

 significant differences were found when the 

 differences among samples were taken into 

 consideration in the Approximate F Test (Ap- 

 pendix Table A5). As the years tested con- 

 tained time differentials from 1 to 22 years, 

 both short- and long-term changes appear 

 nonsignificant. 



Areas 



Comparisons were made between samples 

 from eastern and western Georges Bank within 

 year, month, and size category strata in the 

 same manner as described above. No significant 

 differences were found when the Approximate 

 F Test using sample-to -sample differences was 

 applied (Appendix Table A6). 



The same procedure was followed to test 





