FISHERY BULLETIN: VOL. 86. NO. 1 



Figure 6. — Yellowtail flounder fishing grounds defined by U.S. statistical area are as follows: Southern New 

 England. 526-539; Georges Bank, 522-525; Cape Cod, 514 and 521. 



3 (Table lb) were also selected from six possible 

 zones based on the bathymetric distribution of 

 yellowtail flounder. 



Trip data were aggi'egated at different levels of 

 spatial resolution to examine variability in 

 CPUE over the entire region and within each of 

 the three established stocks. Two-way ANOVAs 

 with interaction were performed on annual data 

 sets using the BMDP statistical software progi'am 

 P4V (Dixon 1981). Given the large number of ob- 

 servations in each analysis, the more rigorous 

 99% significance level was chosen to test the null 

 hypothesis (no significant differences) since 

 relatively small differences in mean CPUE can 

 produce statistically significant results. The 

 ANOVA was performed initially to test for differ- 

 ences in CPUE among all tonnage classes and 

 statistical areas and to determine the overall ex- 

 tent of tonnage class-area interactions. Second- 

 ary analyses were performed to examine the ef- 

 fect of tonnage class-area and tonnage class-depth 

 interactions among and within each of the stocks. 

 All subsequent tests for significance of tonnage 

 class, area, and depth main effects were per- 

 formed with the interaction effects absorbed in 

 the error sum of squares. Estimates of annual 

 geometric mean CPUE were obtained by combi- 

 nations of tonnage class, stock area, and depth 

 from the row and column means provided by the 

 P4V software (Tables 2, 3). 



A standard vessel class was selected for each of 

 the three stocks for use in calculation of fishing 

 power coefficients based on the prevalence of the 

 vessel class in the fishery and its relative contri- 

 bution to the landings over the 20 years. The TC 

 32 category was chosen as the standard for both 

 Georges Bank and Southern New England stocks, 

 and the TC 25 class was chosen as the standard 

 for the Cape Cod stock. Within each stock annual 

 fishing power coefficients were derived for each 

 tonnage class relative to the standard by fitting 

 In CPUE to a one-way linear model using the 

 GLM procedure of the Statistical Analysis Sys- 

 tem (SAS Institute 1982) as follows: 



U = CC + ^ [^jXj] + ^. 



Annual deviations of the coefficients from the 

 20-yr mean were tested for first order autocorre- 

 lation using the Durbin-Watson test statistic 

 (Neter and Wasserman 1974). A time component 

 was subsequently incorporated in the linear 

 model to account for annual trends in the data; 

 seasonal effects were also included by classifying 

 the data according to calendar quarter. The ini- 

 tial year (1964) and the fourth quarter were se- 

 lected as reference categories. The general model 

 is specified as: 



98 



