828 



Fishery Bulletin 100(4) 



1 



08 

 06 



0,4 



02 







12 

 1 



0.8 

 2 0,6 



1 0,4 



i 02 ^ 

 o 



1 5 

 1 



05 

 



2 

 1 5 



1 

 05 







Female 



Age 2 



+ i-tf't^jH^ 



JFMAMJJASOND JFMAMJJASOND 



Age 3 



» 4 6 



1 



• r^'t^ t 



_i I I I 'III 



J FMAMJ J ASOND 



V " 1 



Age 4 



HJ 



J FMAMJ J ASOND 



1 2 1 3 

 ^ . . . 2 



Age 5 



1 



08 

 06 

 04 

 02 







1 5 

 1 



05 

 



1 5 



2 



J FMAMJ J ASOND 





4 6 1 



J FMAMJ J ASOND 



1-5 •2 



05 1- 

 



*■ T 4 



\T 4 , 1 



J FMAMJ J ASOND 



J FMAMJ J ASOND 



Month 



Figure 8 



Monthly means of marginal increment ratio of female and male swordfish 

 in the waters around Taiwan for ages 2 to 5 respectively. Vertical bars are 1 

 SE, numbers on the top of vertical bars are sample sizes. 



curves for males and females with method I and method 

 II back-calculation are shown in Figure 9, and the esti- 

 mated parameters corresponding to each curve are shown 

 in Table 3. 



Hotelling's T- test results showed significant difference 

 in growth parameters between female and male swordfish 

 for either standard or generalized VB with either method I 

 or method II back-calculation (Table 4). The calculated T~ 

 is considerably higher than the tabulated value in Table 

 4 for each case, and all the parameters, except for m of 

 the generalized VB, significantly affect the differences in 

 growth between male and female swordfish (The Roy-Bose 

 simultaneous confidence intervals around differences be- 

 tween parameter values fail to include zeroi. The results 

 of goodness-of-fit comparison showed that the generalized 

 VB had larger r~ ranks for method II, but had equal ranks 

 with the standard VB for method I. Considering the tie 

 rank groups in each sex-by-method, Friedman's x^^ sta- 

 tistics is 2 (n=4). This was not significant at the 5')?^ level 

 (i.e. ;|f-=3.841, df=l). which indicated no significant differ- 



ence in the r- rank ordering of the values between the two 

 growth functions. Kendall's coefficient of concordance was 

 0.5 (;i=4), which did not indicate a good agreement in the 

 r- ranks for all sexes-by-method. 



Discussion 



Just as reported by Berkeley and Houde (1983), Radtke 

 and Hurley (1983), Wilson and Dean (1983), Tsimenides 

 and Tserpes (1989), Ehrhardt (1992), Tserpes and Tsi- 

 menides (1995), and Stone and Porter (1997), females in 

 our study were typically larger than males although the 

 length-weight relationship between the sexes did not 

 differ significantly. The overall sex ratio for the sampling 

 period in our study did not deviate significantly from 1: 

 1 (P<0.01) but differed substantially from the ratios of 

 2.3 females to 1 male and 2.7 females to 1 male reported 

 respectively by Stone and Porter (1997) and Caton et al. 

 ( 1998). This discrepancy may have been caused by the dif- 



