Van Eeckhaute et al ; Movements of Melanogrammus aeglefinus determined from a population model 



665 



Treatment of missing observations is summarized in 

 Table 3. 



The multiplicative model requires a In transfor- 

 mation of the catch per tow from the surveys. This 

 necessitated some treatment of remaining "0" val- 

 ues in the data. To reduce the influence of "0" val- 

 ues, the strata means rather than individual tow data 

 were used because there are frequent occurrences of 

 "0" catch in the individual tow data. Consequently, a 

 test for the significance of interaction between years 

 and strata was not possible. Because a logarithmic 

 transformation of the mean catch per tow is required, 

 a constant less than the minimum catch was added 

 to "0" values. To examine the impact of adding a con- 

 stant to the zero observations, analyses using three 

 different constants, 0.1, 0.01, and 0.001, were con- 



sidered. It was found that smaller constants exerted 

 a greater influence on the model's output as seen by 

 increases in Cook's D statistic (Cook, 1977). More- 

 over, the frequency distribution of the residuals 

 showed greater departure from a normal distribu- 

 tion when 0.01 was added and particularly when 

 0.001 was added, displaying a marked skew toward 

 negative values. Consequently, 0. 1 was added to zero 

 values for multiplicative model computations. The 

 smallest observed value greater than "0" was 0.2. 



For each survey, an age-length key combining all 

 strata on eastern Georges Bank was applied to the 

 length composition in each stratum or stratum sec- 

 tion to derive the age structure. The age structures 

 to which the estimated mean number per tow for 

 missing values were applied were taken from the 



