720 



Fishery Bulletin 92(4). 1994 



Sensitivity to weights 



We varied weights (A ( ) in the objective function of 

 the SMPAR model to determine how sensitive recent 

 biomass estimates were to different types of data and 

 to the recruitment constraint. Biomass estimates for 

 age 1+ northern anchovy during the 1991 fishing 

 season and average biomass during the 1985 to 1991 

 fishing seasons were calculated with the weight for 

 each data type set equal to a range of values while 

 other weights were kept at 1.0. Sensitivity analysis 

 for the weight applied to the recruitment constraint 

 (A 6 in Eqn. 11) is equivalent to a sensitivity analysis 

 on the standard deviation for log-scale recruitments 

 (o in Eqn. 13). Doubling the weight (A 6 =2.0) is the 

 same as reducing the assumed standard deviation 

 by 1/V2=0.71. For simplicity biomass levels in sensi- 

 tivity analyses were not corrected for bias. SMPAR 

 did not completely converge when the weight for the 

 recruitment constraint was set to 0.0 or 10.0. 



Results indicate that current biomass estimates 

 and management advice are affected by weights on 

 EPI data and the recruitment constraint (Table 5). 

 Estimates of anchovy biomass for 1991 were sensi- 

 tive (changes >10%) to halving (A,=0.5) or doubling 

 (A,=2.0) weights on EPI data and the recruitment 

 constraint, but estimates of average biomass during 

 1985 to 1991 were not. Biomass estimates for 1991 

 were sensitive to removing (A ( =0.0) EPI data, EPM 

 data, or the recruitment constraint from the model. 

 Average biomass estimates were sensitive to remov- 

 ing EPI, EPM, or SPOTTER data. 



Simulation analyses 



We used simulation analyses to determine how well 

 SMPAR would estimate biomass under a wide range 

 of recruitment and fishing mortality rates during the 

 1985 to 1991 fishing seasons. Calculations were the 

 same as those for bootstrapping except that fishing 



