FISHERY BULLETIN: VOL. 84, NO. 4 



a standard procedure to plot empirical values of K x 

 obtained against the corresponding body weights, 

 i.e., the mean weights (W) corresponding to each 

 growth increment, or 



logio Ki = log 10 a + b log 10 W 

 which leads to the model 



K x = aW b . 



(2) 



(3) 



(See Sprugel 1983 for a method to correct the bias 

 due to log transformation in this and the other 

 models below.) A discussion of this model may be 

 found in Jones (1976) (see Figure la for an example). 

 This model has three liabilities, the first of which 

 is the most serious: 



1) The parameters "a" and "6" have no biological 

 meaning, i.e., cannot be predicted from one's 

 knowledge of the biology of a given fish. Converse- 



Traditional model 

 (r 2 = 0.821) 





New model 

 (r 2 = 0.888) 



,oq io w °> 



2 3 



Body weight (g,log |0 units) 



ooi 



o 



.s> i.o 



<u 



c 

 o 



w 



a) 



> 

 c 

 o 

 o 



0.6 



0.4 



0.2 



Traditional 

 model 



Upper limit 

 for new model 

 (K=l,when W=0) 



— Traditional model 



— New model 



Lower limit 



for new model Traditional 



(KpO when W» Woo) model 



\ 



50 



100 



_i ty\/\ s---ii- , 



150 1,500 



00 



Body weight (g) 



Figure 1.— Relationship of gross food conversion efficiency (K{) and body weight (W) in Channa striata, a) 

 Plot of \og w K 1 on log 10 W^, as needed to estimate parameters "a" and "b" of traditional model for prediction 

 of K x from body weight, b) Plot of -log 10 (l -K x ) on log 10 W, as needed to estimate parameters W m and p of 

 new model, c) Comparison of the two models. Note that both fit the data well over the range for which data 

 points are available, but that the traditional model provides nonsensical results beyond this range (see text). 

 Based on the data in Pandian (1967). 



828 



