722 



Fishery Bulletin 100(4) 



00 

 <D 



m 



•■a- 



CO 



B 



in 

 d 



in 

 d 



11 16 24 36 



59 



88 



102 



119 



Haul 



1 11 16 24 



36 



59 



88 



102 



119 



Haul 



Figure 1 



I A) Catches v, ^ from the parallel trawling experiments in Table 1. The groups are separated by vertical lines and the 

 group numbers corresponding to Table 1 are given at the bottom of the figure. The symbols represent vessels (see Table 

 1), and the cumulative number of hauls are given at the horizontal axis. (B) Corresponding differences z, =,v, i - v, ■>■ The 

 symbols indicate gear type; 3236: Campelen 1800 shrimp trawl with 35-mm mesh size, 40 m sweeps and rockhopper 

 gear; 3270: same as 3236, but 20-mm mesh size; 3271: same as 3270 but with strapping. 3270/1 indicate that one vessel 

 uses 3270, the other 3271. To the right, 95'?r confidence intervals for the group means are given. The first and second 

 vessel in each group are given at the top and bottom, respectively, for each group. 



but as an approximation they will be considered identical. 

 However, we will allow for an additive individual vessel 

 effect a,, j=l,2, which permits differences in equipment 

 and efficiency for the two vessels. This leads to the fol- 

 lowing model for the observations [v,^, i=\, ... , /!;./=l,2) 

 where j=l, 2 corresponds to the first and second vessel in 

 Table 1: 



.V,;l=A-^,l -V' + ^l+^^l 



.>':,2 = /'^i xJ + o-1^^,.2- 



(1) 



All the factors affecting jointly tow performance are 

 supposed to be in the function /", and therefore the residuals 

 |e, , i=\ n;j=l,2] are assumed to be independent zero- 

 mean identically distributed random variables and a^ = sd 

 (£,-j) is the measurement error. 



We can now eliminate /' and the explanatory variables 

 by taking differences, i.e. 



z, = y, 



«l-«2 + ^,;l-^,;2- 



The expected difference between the two vessels is then 

 given by 



E(2, ) = a, - Oj 

 and because of the independence off, j and e^.^, 

 o\ = var(2,) = var(f, J -f,_2' = 2(7^, 

 and the standard error CT can be estimated as 



- 1 - 1 1 V, 



(2) 



