FISHERY BULLETIN VOL 77. NO 



According to the null hypothesis, the laboratory 

 growth curve corresponding to the temperature at 

 which the sample was taken should provide the 

 best fit to the data. A suitable parametric proce- 

 dure for testing this goodness of fit was not avail- 

 able. The nonlinear least-squares method for 

 fitting the Gompertz growth function to size at age 

 data (Zweifel and Lasker 1976) does provide esti- 

 mates of confidence intervals on the parameters, 

 but the probability levels associated with these 

 intervals are only approximate (Conway et al. 



1970). We found these confidence intervals to be 

 rather broad. 



We tested the goodness of fit by using a modifica- 

 tion of the median regression procedure of Tate 

 and Clelland ( 1957), replacing their estimated re- 

 gression line by the laboratory derived Gompertz 

 growth curve specified by our null hypothesis. 

 Only curves at 1' C intervals were considered. 

 Each larva in a sample was classified into a 2 x 2 

 contingency table, according to whether it was to 

 the left or right of the median age of larvae in the 



Al 

 I5 8°C 



_!_, 1 l_l I i I U 



15 l-C 



II II 



NUMBER OF INCREMENTS 



NUMBER OF INCREMENTS 



416 



