FISHERY BULLETIN: VOL. 80, NO. 3 



could further confound the interpretation of 

 effort statistics. It is unknown whether the 

 percentage of reported catch to total catch, 

 among both commercial and recreational 

 fishermen, is increasing, decreasing, or remain- 

 ing stable. There has also been a trend toward in- 

 creased fishing power with the advent of 

 mechanical line haulers, but the exact amount of 

 this effect is unknown. Considerations such as 

 these make it difficult to quantify bottom fish 

 effort statistics. 



This brief discussion underscores the impor- 

 tance of employing an appropriate measure of 

 fishing effort in which catchability does not vary 

 according to the activities of man. It was possible 

 to demonstrate the superiority of fisherman- 

 days over catch records and yet the former 

 measure proved to be inadequate when pooling 

 across banks was attempted. 



Effects of Aggregation 



Investigators have reported that in a multi- 

 species fishery the TBSM when applied to aggre- 

 gated data often fits better than the Schaefer 

 model applied on a species-by-species basis 

 (FAO 1978; Pauly 1979; Pope 1979). We will 

 examine this phenomenon for data from the 

 MLKM bank for two levels of aggregation. 



As shown in the results section we have applied 

 the Schaefer model to CPUE and effort data at 

 three levels of data aggregation. First we 

 applied the Schaefer model to the data on a 

 species-by-species basis. Then species were 

 partitioned into three cluster groups, the catch 

 and effort data were computed for each group, 

 and the TBSM was fitted to each group. Finally 

 all species were pooled into one group and the 

 aggregate data consisting of total catch and 

 effort were computed and fitted with the TBSM. 

 The fit of the TBSM to each of the three species 

 groups and to the total group resulted in 

 significant regressions for the MLKM bank 

 while only 2 out of 13 single-species regressions 

 for this bank were significant. This result may be 

 due to the fact that the fishery exploits groups of 

 species simultaneously and that our measure of 

 fishing effort measures exploitation on species 

 groups rather than single species. It is apparent 

 that when the data in this study were progres- 

 sively pooled, the correlation coefficients 

 describing the fit became increasingly negative 

 (Fig. 4). This result alone would suggest that 

 aggregation led to a better fit. Unfortunately be- 



>- 



o 



o 



a. 



2 



I 



r~ 





! — f" 



1 



13 INDEPENDENT SPECIES 

 (n-52) 



LL 



n 



n ljj 



3 CLUSTER ANALYSIS SPECIES GROUPS 

 (n-12) 



_□ . □ 



J L 



n n n 



TOTAL AGGREGATE 

 (n = 4) 



"4 -.2 +.2 + .4 +.6 



CORRELATION COEFFICIENT (r) 



+.8 



Figure 4.— Frequency distributions of correlation of co- 

 efficients between CPUE and fishing effort based on three 

 levels of species aggregation. 



cause only 2 out of the 13 single-species regres- 

 sions were signficant, it is not appropriate to use 

 the single-species results in our comparison of 

 the effects of aggregation. Table 4 presents the 

 correlation coefficients between CPUE and 

 effort for each of the three cluster groups and the 

 total aggregate. At first glance it appears that 

 for the MLKM bank the TBSM applied to the 

 total group fits substantially better (r 2 = 0.77 for 

 fisherman-days) than the TBSM applied to any 

 of the three species groups (r 2 = 0.25, r 2 = 0.59, 

 and r 2 = 0.25). However, an examination of the 

 correlations between fishing effort for the three 

 cluster groups reveals that these variables are 

 highly correlated (Table 8). Grunfeld and 

 Griliches (1960) have cogently argued that 

 increased colinearity of independent variables 

 can lead to an increase in the goodness of fit (r 2 ) 

 when data have been aggregated. This deceptive 

 gain in the explanatory power of an aggregated 

 independent variable prevents a direct compari- 



Table 8.— Correlations of fishing effort (fisherman- 

 days) if) among cluster analysis species groups. 



Group 

 effort 



12 



13 



n 



f2 



1.000 



0.943* 

 1.000 



0.900* 

 0.940* 

 1.000 



'Significant P = 0.01. df =78. 



444 



