Paired Models are often used to examine the relative effectiveness of alternative 



Comparisons management proposals for increasing fish production. The effects of man- 



agement changes could be lost or confounded with random fluctuations in- 

 troduced by stochastic models such as the SLCM. To facilitate meaningful 

 comparisons of alternative management proposals, the SLCM is designed 

 to allow the user to control the random number streams used within the 

 model. This allows users to replicate serial patterns within the SLCM re- 

 flecting changes in environmental conditions over time. Thus, a user could 

 compare two or more management alternatives over the same temporal 

 patterns in survival. For example, figure 10 shows two time series for a 

 matched pair of games where downstream migrant survival has been in- 

 creased by 40 percent in the second game. While there is independent 

 variation in each game, a significant portion of the variation is shared. 

 Multiple paired comparisons can lead to distributions of comparison statis- 

 tics in much the same way that multiple games lead to distributions of 

 state variables. 



SENSITIVITY ANALYSIS 



The structure of the SLCM allows explicit inclusion of parameter uncer- 

 tainty in the model. By increasing the variance of survival processes, users 

 can include some of the existing uncertainty about survival or production 

 parameters. But increasing the variance only addresses questions of preci- 

 sion; it does not directly address problems of accuracy that arise if the pa- 

 rameter estimates are biased. A rigorous sensitivity analysis is needed to 

 assess the ramifications of potential bias in parameter estimates. 



The SLCM's relatively straightforward structure allows sensitivity 

 analyses to be performed easily. The model can be reduced to five major 

 components: density-independent juvenile production and survival, 



12,000 

 11,000 

 10,000 



1,000 ' ' 



10 20 30 40 50 60 70 80 90 100 



Figure 10 — Paired comparison of model results using two levels 

 of passage survival: 32 percent (lower line) and 44.8 percent 

 (upper line). 



18 



