632 



Fishery Bulletin 93(4). 1995 



were used to compare mean size and age of leopard 

 coralgrouper on the four reefs, independent of the 

 reef fishing status. Multiple comparisons were per- 

 formed by using post hoc tests (Tukey-Kramer, level 

 of significance P=<0.05), and pair- wise comparisons 

 by using Kolmogorov-Smirnov tests. Schnute's 

 growth function ( 1981) was used to fit length-at-age 

 data of leopard coralgrouper for each reef by using 

 standard nonlinear optimization methods (Wilkinson, 

 1989). Schnute's model includes the von Bertalanffy, 

 Richards, Gompertz, logistic, and linear growth mod- 

 els, which correspond simply to limiting parameter 

 values. To test for differences in size at age between 

 reefs, the linear regressions were compared by us- 

 ing analysis of covariance. Chi-square contingency 

 tables were used to compare the frequency of sexes 

 between reefs. The assumptions of normality and 

 homoscedasticity were examined and data were 

 transformed if needed (transformed data are indi- 

 cated in tables). Level of significance used was 

 P<0.05. 



Results 



There were no significant differences in mean size 

 and age between protected reefs and unprotected 

 reefs (fishing status). However, the mean sizes and 

 ages varied significantly between reefs within fish- 

 ing status level (Table 2). 



Post hoc tests showed that mean size and mean 

 age were larger for Glow (closed) than for all other 

 reefs, whereas mean ages for Grub (open) were 

 smaller than for all other reefs. The mean sizes were 

 not significantly different for Yankee, Hopkinson, and 

 Grub, and the mean ages were not significantly dif- 

 ferent for Yankee and Hopkinson (Fig. 2). 



Growth 



Schnute's growth function was fitted to size-at-age 

 data for each reef. The submodel corresponding to 

 the von Bertalanffy formula (b=l) provided a good 

 fit to the data from all reefs (Fig. 3). The estimated 

 "a" ( corresponding to the von Bertalanffy K) for Grub, 

 however, approached zero, indicating that the data 

 could be described also by a linear regression model. 

 Ferreira and Russ (1994) found that for the leop- 

 ard coralgrouper, estimates of growth parameters are 

 affected greatly by different age ranges of size-at- 

 age data. Therefore, for comparison of growth be- 

 tween reefs, the age range was limited to age classes 

 occurring at all four reefs (2 to 10 years), and 

 Schnute's growth function was fitted to these trun- 

 cated data. For Hopkinson, Grub, and Glow, esti- 

 mates of "a" approached zero (Table 3), indicating 

 close to linear growth over the age range 2 to 10 years. 

 As the estimate of "a" for Yankee was also low, simple 

 linear models were fitted to the data from all four 

 reefs for comparative purposes (Table 3). Analysis of 

 the sum of squares indicated that linear models were 



Table 2 



Nested analysis of variance comparing mean size and age of leopard coralgrouper, Plectropomus leopardus, from reefs open and 

 closed to fishing. The difference between residual degrees of freedom in the two tables is due to the fact that for some individuals 

 age was not determined, df = degrees of freedom; SS = sum of squares; MS = mean square. 



Source 



df 



SS 



MS 



F-value 



Dependent variable: FL (cm) 



Fishing status 1 



Reef (fishing status) 2 



Residual 443 



Dependent variable: Log age (years) 



Fishing status 1 



Reef (fishing status) 2 



Residual 413 



393.6 

 531.1 



18870.6 



0.931 



0.422 



9.61 



393.6 



265.5 



42.6 



0.931 

 0.211 

 0.023 



1.48 

 6.23 



4.41 

 9.07 



P-value 



0.35 

 0.002 



0.17 

 0.0001 



