0.30, so that F+M remained 0.55. A further ad- 

 justment was also made in the recruitment curve 

 (fig. 5C). The third trial, in which these adjust- 

 ments were employed, yielded a further improve- 

 ment in the fit (r = 0.69), which is shown in figure 6. 

 A notable feature of the trials was the sensitivity of 

 the fits to changes in the stock-recruitment relation. 

 The greatest improvement in fit occurred between 

 the first two trials, as the result solely of a moderate 

 change in the shape of the recruit curve (figs. 5A, 5B). 

 The series Yu- and Y^ are not, of course, com- 

 pletely independent, since F,„ enters to some extent 

 into the calculation of the parameters used to com- 

 pute y,r. As a relative measure of goodness of fit, 

 however, the coefficient is of some value. It is 

 possible to make at least two positive statements 

 regarding r as used here : 



1 . If the calculated P value is greater than the 

 significance level, the real value is certainly 



greater. Additional degrees of freedom will 

 be lost according to the degree of dependence 

 of the variables. It is possible, then, to 

 identify markedly nonsignificant fits. 



2. For calculated correlations within conven- 

 tional significance levels, the value of r serves 

 as a means of comparing two fits. If one set 

 of parameters generates a fit with a higher 

 value of r than another, then it should be 

 closer to the truth than the other. 



The parameters used in the simulations have vary- 

 ing and subtle degrees of dependency on the catch 

 data. Thus, to assess accurately the value of P for 

 the coefficients would require a lengthy and com- 

 plicated statistical analysis. This work appears 

 hardly to be warranted in view of the approximate 

 nature of the simulations. 



] 932 1934 193 6 1938 1940 1942 1944 1946 1948 1950 1952 1954 1956 1958 



FISHING SEASON 



Figure 6.— Population and yield simulation for New England cod; M = 0.25, C! = 1.47, and g = 0.34 for entire period. 



Other parameters as shown in drawing. 



ANALOG COMPUTER MODELS OF FISH POPULATIONS 



41 



