Chen and Montgomery: Modeling the dynamics of Jasus verreauxi 



33 



ish seiners) that operated from 

 lobster producing ports and in 

 oceanic waters was used as an 

 index of fishing effort for the 

 lobster fishery. This was done 

 because it was impossible to 

 distinguish between vessels 

 that were used for fishing lob- 

 sters and vessels that were 

 used for fishing other species 

 in the same area (Montgomery, 

 1995). During this period, an- 

 nual landings of lobsters were 

 probably under-reported be- 

 cause only lobsters sold at the 

 main market centers were re- 

 corded. Such an over-estima- 

 tion of efforts and under-esti- 

 mation of catches would lead 

 to under-estimation of CPUE 

 for this period of time. It 

 should be realized, however, 

 that the above reasons for the 

 large deviation of observed 

 data from model estimates re- 

 main a working hypothesis. 



Outliers can be identified in 

 linear regression analyses by 

 using criteria developed by 

 Rousseeuw and Leroy (1987). 

 However, the criteria devel- 

 oped through extensive simu- 

 lations based on a linear re- 

 gression model cannot be used 

 in nonlinear regression analy- 

 sis. An extensive simulation 

 study is needed to develop suit- 

 able criteria to identify outli- 

 ers in nonlinear regression 

 analyses. Before this can be 

 done, the LMSE-based re- 

 weighted least squares, which 

 has been shown to behave bet- 

 ter than the least median of 

 squares in linear regression 

 analyses (Chen et al., 1994; 

 Chen and Paloheimo, 1995), 

 cannot be used in nonlinear regression analyses. 



The results presented in our paper were based on 

 catch data adjusted by using one set of adjustment 

 coefficients. To test the sensitivity of the results to 

 the adjustment coefficients, we also used two other 

 sets of adjustment coefficients in the analysis (Table 

 4). These two sets of coefficients have more extreme 

 values than the one presented in Table 1. The detailed 



results for these two sets of data were reported in Mont- 

 gomery and Chen ( 1996). We plotted the stock biomass 

 estimated with these three sets of data (Fig. 10). Al- 

 though there are differences, they are rather small con- 

 sidering the large differences in adjustment coefficients. 

 This may indicate that the results presented in this 

 paper tend not to be sensitive to small changes in catch 

 adjustment coefficients used in period II. 



