Laake et al.: Pinniped diet composition 



447 



^Var{^ 



Var(w^j) = 6l 



lation correction (fpc) factor on the second term. If we had 

 used the fpc for salmonids, we would have had to replace 

 the unknown «■ with their estimated values. 



For variance estimates of number of prey consumed (P- ) 

 we also used a delta method approximation: 



where the variance of li' ,^ is the prediction variance lor the 

 ''' otolith. When these data were analyzed, the available 



u 



repression equations in a draft version of Hai"vey et al. 

 (2000) (lid not include standard errors for the parameters 

 nor the residual variance that were needed to compute the 

 prediction variance. Therefore, unfortunately we had to 

 drop the first term from the variance. The necessary values 

 are now available in Hai-vcy et al. (2000). The second-term 

 measures variability between prey and is typically larger 

 than the prediction error. However, to avoid underestimat- 

 ing this component of variance we dropped the finite popu- 



Var{P,j ) = P,J[cu''{^j ) -I- cv'{p,j )J. 



We did not have variance estimates for energetic require- 

 ments nor age structure; therefore the variation in 

 estimated prey biomass requirement (£ ) reflected only 

 variation in our estimates of population size: 



evil 



-cviN,} 



,jcv^{Cj) + cv^{n, 



which was also approximated by the delta method. 



