FISHERY BULLETIN: VOL. 74, NO. 1 



Differences in Log Catch Per Effective 



Trip Between Vessel Classes, 



Between Areas, and Among Years 



A factorial analysis of variance in a ran- 

 domized complete-block design was used to test 

 whether significant differences occurred in log 

 CIET between vessel classes (blocks), and be- 

 tween areas and among years (main treatment 

 effects). The analysis showed that log CIET with 

 respect to the two vessel classes differed sig- 

 nificantly {F = 12.34; df = 1 and 265; P<0.01). 

 Significant differences in log CIET also occurred 

 with respect to inshore and offshore areas fished 

 {F = 9.38; df = 1 and 5;P<0.05). Furthermore, the 

 results showed significant differences occurred 

 among years fished {F = 9.45; df = 5 and 5; 

 P<0.05). A Duncan multiple-range test (Steel 

 and Torrie 1960), wath Kramer's (1956) extension 

 of the test, determined that a significant differ- 

 ence in the means occurred primarily between 

 1965 and 1969, years in which there were consid- 

 erable differences in fishing conditions. 



Relation Between Log Catch Per Day 

 Fished and Log Catch Per Effective Trip 



Log CIDF increased linearly with log CIET in 

 each of the areas within the size classes. Regres- 

 sion lines, fitted to the data pooled for 1965-70, 

 showed that the scatter about the regression lines 

 was relatively narrow; there were, however, a few 

 observations in each set of data that appeared to 

 have large residuals. To assess the validity or ap- 

 propriateness of the least-squares fitting of log 

 CIDF on log CIET, these residuals were analyzed. 



Figure 4 shows the scatter diagrams in which 

 the residuals were plotted against log CIET for 

 the four sets of data. With the exception of a few 

 outliers which can be seen as isolated points with 

 extreme negative ordinates, there were no 

 noticeable peculiarities in the distribution of the 

 residuals. The outliers were rejected at a multiple 

 of the standard deviation using a premium of 

 2.5% (see Anscombe and Tukey 1963). The 

 overall distribution of the residuals after the 

 rejection procedure appeared in the form of a 

 horizontal band, which indicated that the least- 

 squares analysis of the log transformed data was 

 satisfactory. 



After the rejection of large residuals, regres- 

 sion lines were fitted to the data as shown in Fig- 

 ure 5. The dashed lines on either side of the re- 



+0 3 



+0.2 



+0.1 







-0.1 



-02 



-03 

 +0 3 



+0 2 



+0 1 



3 -0 2 

 I 



CLASS I (INSHORE) 



REJECTED 



I I I I I I I L 



>s 



-03 



-0 4 

 g +03 



- +0 2 



< +0.1 

 o 



Ui 



"= -0 I h 



-02 



-03 

 +03 



+0 2 



+0 I 







-0.1 



-0.2 



-0.3 



-QA 



-0.4 -02 +0 2 +0 4 +0 6 +0 8 



LOG C/ET 



FIGURE 4.— Plots of residuals (log C/DF - log CrDF) against 

 log C/ET for class 1 and class 2 vessels fishing inshore and 

 offshore in 1965-70. 



gression lines indicate the 95% confidence limits 

 for the estimates of log CIDF. The values of the 

 regression equation and correlation coefficient of 

 log CIDF on log CIET are given in Table 3. 



Substitution of values of log a and b into the 

 logarithmic equation logioC/DF = logioa + 

 blogioC lET and solution of the equation provided 

 estimates of CIDF from CIET, by month, for 



Table 3. — Data on the regression and correlation of 

 \ogioC/DF on logioC/£r in the Hawaiian skipjack tuna fishery, 

 by vessel size classes and areas, 1965-70. Two asterisks 

 denote probabilities equal to or less than 0.01. 



64 



