584 



Fishery Bulletin 89|4], 1991 



tions. The effect of these differences on the mean and 

 standard deviation of length-at-age is shown in Table 

 2. Despite the differences in parameter estimates, 

 mean length-at-age is virtually identical for all models 

 up to age 7. This allows the calculation of an approx- 

 imate age-at-release for each observation that should 

 contain little error and that is essentially independent 

 of which set of parameter estimates is used to convert 

 the release length to age. With this information, a 

 presumed age-at-recapture incorporating individual 

 variation can be calculated by adding the presumed 

 age-at-release to the observed period at liberty. A plot 

 of presumed age-at-recapture (calculated using the 

 model 3 parameter estimates) against recapture length 

 is shown in Figure 3. 



Model 1 produced the highest estimate of \* L and 

 the lowest estimate of \i K . The probable explanation 

 for this is that the estimates are biased by the three 

 observations with the greatest recapture lengths (Fig. 

 3). Because there is no individual variation in this 

 model, the estimate of L^ is forced upward to compen- 

 sate for these influential observations. In model 2, such 

 observations can be accommodated by individual varia- 

 tion in L^; thus the estimate of n L ^ falls by more 

 than 20% and the estimate of ^ K rises by more than 

 50%. When a model error term is added (model 3), the 

 differences in relation to model 1 are less dramatic. The 

 addition of release-length-measurement error (model 



4) has little effect on the model-3 estimates, with the 

 exception of o e 2 , which is somewhat lower in model 4. 

 Similar effects of adding model error and release- 

 length-measurement error can be seen in comparisons 

 of the parameter estimates for models 5, 6, and 7. 



The estimates of o L 2 are substantially higher for 

 models that assume constant K (models 2,-3, and 4) 

 than for those that assume individual variability in K 



