296 



Fishery Bulletin 101(2) 



17 



16 



15 



14 



13 



12 



11 



10 



9 



8 



7 



6 



5 



4 



3 



2 



1 



00 



—I 1 r 



0,2 0.4 0,6 



Mean CV by species 



0,8 



Figure 2 



Mean CVs, by species, of biomass indices in the trawl survey data sets. 

 Each plotted point relates to one species in a trawl survey series, and 

 indicates the mean of all estimated CVs for that species in that series. 

 Points are jittered vertically to avoid overlap. 



Table 1 



Three alternative error-distribution assumptions for biomass indices in stock assessments, and the associated form of the stan- 

 dardized residuals. Notation: /, is the ith biomass index and B, is the corresponding model estimate of (absolute) biomass; for 

 assumption Inorm, 0'f = log (c^ + 1). 



Label 



Description 



Standardized residual 



norm /, is normally distributed with mean qB^ and assumed CV c, 



Inorm / is lognormally distributed with mean gS, and assumed CV c, 



lognorm log(/, ) is normally distributed with mean logl^B, ) and s.d. c, 



c, 1, f/B, 



log 



qB.) 



-I- 0.5(1^ 



— log -^ 



Are the assessment CVs the right size? 



We constructed a residual statistic, V, that was designed 

 to indicate whether the CVs assumed in the stock assess- 

 ment (the t,) were too small or too large in each data set. A 

 positive (or negative) value of ^suggests that the residuals 



were too large (too small), and thus CVs were too small 

 (too large). The statistic is based on the median absolute 

 standardized residual (MASK), rather than the residual 

 variance, b(>cause the latter is not very robust (it is easily 

 inflated by outliers). We defined 



