Dressel and Norcross: Using poststrafication to improve abundance estimates from multispeaes surveys 



477 



Calculations of relative efficiency among the three to- 

 tal abundance estimators showed increases in estimated 

 precision with stratification (Table 5). In most cases (18 

 out of 24), the estimate poststratified by habitat was 

 more precise (corresponding to a lower standard error 

 in Fig. 5) than the unstratified estimate. Of the 16 (of 

 23) cases in which the precision of both poststratified 

 estimates were greater than that of the unstratified 

 estimate, in half the estimate poststratified by both 

 habitat and density was more precise than the estima- 

 tor poststratified by habitat alone. 



Sample sizes across the survey area and in each sub- 

 area (habitat, high fish-density, and low fish-density 

 areas) (Table 6) strongly influenced the precision of es- 

 timates. Habitat sample sizes for all species-year combi- 

 nations ranged from 4 to 45 (proportion of samples tak- 

 en in habitat ranged from 0.286 to 1.000); HFD sample 

 sizes ranged from to 29 (proportion of samples taken 

 in the HFD area ranged from 0.0 to 0.8); and LFD 

 sample sizes ranged from 4 to 16 (proportion of samples 

 taken in the LFD area ranged from 0.125 to 0.583). 

 Although the number of samples in both the high and 

 low fish-density areas (Fig. 6, A and B) likely affected 

 estimates poststratified by habitat and fish density, the 

 number of samples in the HFD area appears to have 

 had the primary influence on the precision of estimates. 

 The species-year combinations for which the unstrati- 

 fied estimate was the most precise occurred when habi- 



tat sample sizes ranged from 4 to 22 (Fig. 7) and HFD 

 stratum samples sizes ranged from 6 to 11 (Fig. 6A). 

 The species-year combinations for which the estimate 

 poststratified by habitat was the most precise occurred 

 when habitat sample sizes ranged from 12 to 30 (Fig. 7) 

 and when sample sizes in the HFD stratum ranged 

 from 6 to 15 (Fig. 6A). The species-year combinations 

 for which the estimate poststratified by habitat and fish 

 density was most precise occurred when habitat sample 

 sizes ranged from 15 to 45 (Fig. 7) and HFD stratum 

 sample sizes ranged from 10 to 29 (Fig. 6A). Estimates 

 poststratified by habitat and fish density were the most 

 precise for all three cases in which the HFD stratum 

 sample size was greater than 20 (corresponding to LFD 

 stratum sample sizes ranging from 9 to 16) (Fig. 6, A 

 and B). Both of the poststratified estimates were more 

 precise than the unstratified estimate when habitat 

 stratum sample sizes were greater or equal to 24 (Fig. 

 7) and when HFD stratum sizes were greater or equal 

 to 12 (Fig. 6A). 



Statistically significant changes in annual abundance 

 varied among indices and species. There were signifi- 

 cant changes in annual mean CPUE in all indices for 

 rock sole and Pacific halibut, in two indices for yellowfin 

 sole and in no indices for flathead sole (Table 7). Rock 

 sole abundance was significantly greater in 1992 than 

 all other years except 1996. Individual indices indicated 

 that rock sole 1996 abundance was greater than that 



