shapes and sizes of the tow nets used — a 3-m-diameter 

 round net and a 2.7-m-square net — I made 22 paired 

 tows using the two nets alternately. The sampling con- 

 sisted of 44 standard tows at the surface on four dates 

 in three sampling units (two in Coville Lake and one in 

 Iliuk Arm). The catch of age sockeye salmon for 

 each tow and the figures used to test the hypothesis of 

 no difference in the rate of catch of the two nets with 

 the Wilcoxon matched-pairs signed-ranks test (Siegel, 

 1956) are presented in Table 4. The derived T = 53 

 with N = 19 is greater than the theoretical T = 46 at 

 the 5% level of significance, so the hypothesis of no 

 difference in the effectiveness of the two nets is ac- 

 cepted. The results of sampling with the two types of 

 nets are combined in my subsequent analyses. 



Most of the 1961 tows were made only at the surface 

 and their comparability to the data collected at both 

 depths in later years was uncertain. A split-plot 

 analysis of variance (Snedecor, 1956) was made to de- 

 termine if the average catches of juvenile sockeye 



salmon in surface tows and deep tows were signifi- 

 cantly different. This analysis was restricted to Coville 

 Lake, Iliuk Arm, and South Bay, 1962 to 1964, be- 

 cause of lack of sufficient data from other areas. The 

 model is 



■ijk = ^+ B i + 



M 



+ e U 



+ T + (MT)n. +o 



>jk ^°ijk. 



where X 



ijk 



e U 



Tk 



(MT) jk 



°ijk 



average catch at the Ath depth in the 



jth time period in the z'th area, 



block effect, i.e., area, 



fixed main treatment effect, i.e., time, 



whole plot error, i.e., area-time 



interaction, 



fixed subplot treatment, i.e., depth, 



time-depth interaction, and 



subplot error, i.e., area-depth plus 



area-time depth interactions. 



The split-plot analysis involved only surface and 

 deep paired tows (the subplot treatment), three time 



Table 4. --Numbers of age sockeye salmon taken in 22 paired tows in 3-m-diameter round and 2.7- 

 m-square nets in two sampling units in Coville Lake and one in Iliuk Arm, 1964, and calculations 

 used in Wilcoxon matched-pairs signed-ranks test (Siegel, 19S6) . 



10 



