FISHERY BULLETIN: VOL. 81. NO. 4 



general linear model (Helwig and Council 1979) was 

 used for these analyses, and the partial sums of 

 squares (Type IV) were used as test criteria. 



The von Bertalanffy growth formula was fit to the 

 data from each fish with SAS nonlinear regression, 

 using Marquardfs method (Helwig and Council 

 1979). This provided repeat observations of length at 

 each age, allowing us to estimate variance about the 

 regression line and to compare curves with 

 Hoetelling's T 2 (Morrison 1976; Bernard 1981). 



We divided the fishing grounds into two areas, Hud- 

 son Canyon and Southern New England (Fig. 2), to 

 consider differences in the length frequency of 



tilefish. A weighting procedure was used in calculat- 

 ing length frequencies to eliminate bias introduced 

 by excessively large samples collected during a 

 season. Equal weight was given to each large sample 

 (n >50) from an area in a year and season, and 

 seasonal longline landings (Christensen footnote 5) 

 were used to weight the seasonal length frequencies 

 when computing the annual length frequency. If a 

 year and area had <200 observations, no weighting 

 was used in calculating the length frequencies. 



The total instantaneous mortality rates (Z) were 

 calculated through least-squares regression of the 

 natural log of the number in an age group on age. 



Figure 2.— Map of the continental shelf 

 from New Jersey to Maine showing some of 

 the major submarine canyons from the Mid- 

 dle Atlantic Bight to Georges Bank Two 

 primary fishing areas are labeled, Hudson 

 Canyon (long. 71°55'-72° 42'W) and southern 

 New England (long. 67°00'-71°54'W). 



H yd rog ra pher 



Veatch 

 ■Atlantis 



Middle Grounds 



Block 



\ 



\\ Barnegat Li 



1 



'1 



negat Light 



100 km 



x— : 



50 mi 



754 



