Abstract
This paper presents a new approach to distinguishing an observed length distribution into each of those by age using the genetic algorithms. The length distribution of each age is assumed to follow a normal distribution with mean Tx, and variance σ2x. Three pieces of model independent biological information are incorporated instead of assuming a growth function, and then analyze the undistinguishably mixed length-frequency into each of those by age. The first isthat the number of age class is known. The second and the third are that the ranges of themean length and variance of each age class are roughly known from the other biological studies. For the latter parameter, two models are examined. Artificial data and actual sardine data are analyzed to know the efficiency of this approach. The efficiency depends on the width of ranges given for the mean length and standard deviation of each age class which were used to estimate the parameters mentioned above. However, the results showed even though the width of the ranges werenot so small, both models of this approach worked well to distinguish total length distribution into each of those by age.
