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



483 



Unstratified Poststratified (H) Poststratified (D) 



Most precise estimate of total abundance 



Figure 7 



The habitat stratum sample size for each species-year combination is plotted 

 in relation to the most precise estimate of total abundance — the unstrati- 

 fied estimate, the estimate poststratified by habitat (poststratified [H]), or 

 the estimate poststratified by habitat and fish density (poststratified [D]). 



in the LFD stratum (Fig. 6, A and B). This study sup- 

 ports the conclusion of Scheaffer et al. (1996) but also 

 indicates that the sample size in the HFD stratum may 

 have a larger influence on the precision of the resultant 

 estimate. 



As concluded in other studies (Fiedler and Reilly, 

 1994; Pollock et al., 1994; Reilly and Fiedler, 1994; 



Bernard et al., 1998), we found that poststratification 

 can provide increased precision and decreased bias for 

 estimates. Small stratum sample sizes, however, can 

 make it impossible to detect heterogeneity among strata 

 and fail to give increased precision (Powell et al., 1995; 

 Friedland et al., 1999). The wide range of sample sizes 

 among strata across species-year combinations exempli- 



