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



479 



20,000,000 



15,000,000 



10,000,000 



5,000,000 - 







2,000,000 • 



1,500,000  



1 ,000,000  



500,000 -| 







Rock sole 



DUnstratified 



 PoststraWied(H) 



 Poststratified(D) 



1991 1992 1993 1994 1995 1996 



Yellowfin sole 



fra^foffi 



D Unstratified 

  Poststratified(H) 



IPoststratitied(D) 



2,500,000 

 2,000,000 

 1 .500,000 

 1 ,000,000 

 500,000 

 



1991 1992 1993 1994 1995 1996 



Pacific halibut 



D Unstratified 



 Poststratitied(H) 



 Poststratitied(D) 



1991 1992 1993 1994 1995 1996 



DUnstratified 



 Poststratified(H) 



 Poststratitied(D) 



1991 1992 1993 1994 



Year 



1995 1996 



Figure 4 



Three estimates of total abundance and standard error. Estimates 

 are unstratified, poststratified by habitat (poststratified [H]), and 

 poststratified by habitat and fish density (poststratified [£>]). 



habitat was adjusted by poststratifying by habitat. The 

 estimate poststratified by habitat was less than the 

 unstratified estimate of abundance, as we suspect the 

 true abundance was. Poststratification by habitat and 

 neighboring years' halibut density adjusted not only for 

 the disproportionately large number of samples in the 

 habitat area but also for the disproportionately large 

 number of samples in the HFD area (Table 2). The es- 

 timate poststratified by habitat and halibut density was 

 less than both the estimate poststratified by habitat 

 and the unstratified estimate, as we suspect was the 

 case for the true Pacific halibut abundance. 



In 1992, the number of samples in yellowfin sole 

 habitat was disproportionately large, but the number 



of samples in the HFD area was disproportionately 

 small (Table 2). In this case, we suspect the unstrati- 

 fied estimate of abundance was an overestimate of true 

 abundance because of the overabundance of samples 

 in the habitat area. We also believe, however, that it 

 was not a very large overestimate because of the dis- 

 proportionately small number of samples in the HFD 

 area. Poststratifying by habitat adjusted for the dis- 

 proportionately large number of samples in the habitat 

 area and produced an estimate that was less than the 

 unstratified estimate. Poststratifying by habitat and 

 fish density adjusted for both the disproportionately 

 large number of samples in the habitat area and the 

 disproportionately small number of samples in the HFD 



