FISHERY BI'LLKTIN: VOL. 86, NO, 4 



400000 r- 



g 300000 

 U 



w 

 p- 



2 200000 



w 

 CQ 



IS 



00 



W 



100000 







SMOOTHED 



HLTERED 



PREDICTED 



1930 



1940 



1950 



1960 



1970 



TIME 



Figure 3.— Estimated number of recruits 1939-65, assuming q(a, t) = f(t)s(a). 



to each other, while the predicted values are too 

 smooth. The estimated values of/(f ) (Fig. 4) general- 

 ly decline with time. All three estimates are signifi- 

 cantly different from the previous estimates. Years 

 where relative highs and lows occur differ, show- 

 ing the sensitivity of this class of models to the 



assumed form of the H matrix. Since the estimates 

 oi R, s{a) and/(0 are interrelated, I suspect that 

 the parameter estimates are highly correlated and 

 hence unstable. While I have not calculated the 

 parameter covariance matrix, it should be checked 

 for any serious analysis using these techniques. 



0.4 



^ 0.3 



0.2 



0.1 



H 

 < 



Z 



o 



< 



O 

 J 



CL 



X 



PU 



O 

 H 



< 



Ol, 



pL. 



on 

 U 



0.0 



1930 



1940 



1950 



1960 



1970 



TIME 



Figure 4.— Estimated values of /(O when q(a, t) = f(t)s(a). 



626 



