catch locations was 27 miles and the minimum 

 distance, 3 miles. An analysis of variance 

 (using coded values as outlined by Snedecor, 

 1946) indicated that samples were not drawn 



S 1 8 17 



from a homogeneous population (F = * -- = 



10.38, whereas P 0.05 = ca. 2.84). Four of 

 the catches were separated by distances 

 greater than 12 miles; however, three of the 

 samples (0.38, 0.42, and 0.43) were taken 

 from catches made within three miles of each 

 other. The latter samples showed no statis- 

 tical difference in veiriance. 



Comparisons of samples taken from dif- 

 ferent localities, but landed at the same 

 port, often were found to differ signifi- 

 cantly in length, weight, and age composi- 

 tion. Furthermore, marked variability 

 frequently occurred among sajnples from day 

 to day. In view of these results, the sys- 

 tem of sampling which we eulopted in 1955 

 consisted of tciking daily samples from a 

 large number of schools over the range of 

 the fishery. The number of samples obtained 

 at each port was roughly proportional to 

 both the number of vessels and the catch in 

 each locality. The catches from which the 

 samples were drawn were chosen at random; 

 however, if two or more widely separated 

 fishing areas were represented in the day's 

 catch, a sample was obtained from each. 



Further consideration was given to the 

 number of fish to be measured so as to re- 

 present the size composition of the school 

 from which the sample was drawn. An ex£imple 

 is furnished by a sample of 500 fish drawn 

 from a single school catch (Sample 0.48, 

 Port Monmouth, N. J., September 18, 1952). 

 The vajricUice of the mean, 233 mm., was 117.9. 

 For t .05, n = 20 fish furnishes an approxi- 

 mation of the minimum sample size required 

 for estimating the mean to within 2 percent 

 (Snedecor, 1946). To eliminate bias in the 

 selection of fish for detailed measurements 

 and scale samples, it was decided that 100 

 fish would be chosen to represent a regular 

 sample. From the regular sample, 20 indi- 

 viduals would be selected by laying aside 

 every fifth fish. Results have shown that 

 samples thus collected are not materially 

 biased. 



To determine whether size stratifica- 

 tion occurred in different parts of a vessel 

 load four single-school catches were exam- 

 ined. Five samples, of 100 fish each, were 

 drawn from different locations in the hold 



of each vessel (Appendix table 1). The 

 vessel catches were chosen so as to repre- 

 sent different size groups of fish. An 

 analysis of varisince indicated no signifi- 

 cant difference among the sample means 



140.22 



within vessel catches (F 



= 1.26, 



111.44 



whereas P 0.05 = 1.64). Figure 8 shows the 

 age composition of the individual samples. 

 It is obvious, without recourse to statis- 

 tical procedures, that the vsirious age 

 groups were represented in the same propor- 

 tions throughout the individual samples 

 from a load. These preliminary results have 

 been verified from additional material 

 gathered in other localities along the coast. 



VESSEL I 



V ES S E L 



2 VESSEL 3 VESSEL 4 



12 3 



Figure 8. --Age composition from single-school catches. 



Collection of Samples 



Samples from the purse-seine catches 

 were taken directly from the hold of the 

 vessel. Fish were shoveled into buckets 

 from the top of the load, after first re- 

 moving the sun-dried fish from the upper 

 layer. Each sample thus was presumed 



13 



