664 



Fishery Bulletin 97(3), 1999 



Bank haddock assessment (Gavaris and Van Eeck- 

 haute, 1991 ) and therefore the method used then was 

 followed in our study. For all other missing observa- 

 tions that occurred in strata which did not have a pre- 

 dominance of zero values, a multiplicative model em- 

 ploying strata and years as factors influencing the mean 

 catch per tow was used to derive predicted values: 



In /, , = In /;-,. ^ (In PJX, -h 5^ (In P, )Z, + f , , , 



■s y 



where s = stratum; 



V = year; 



/ = survey catch/tow; 



P - relative power for stratum or year; 



X = dummy variables indicating the stratum 

 and year of observation; and 



e - independent identically distributed nor- 

 mal error. 



Strata sections 17u, 18c, and 18u in the spring and 

 strata sections 17u, 18u, and 19 in the fall were not 

 included in the multiplicative model analysis because 

 of the prevalence of zeros in these strata sections. 



