levels. Typically, to test for differences between arithmetic means, a 

 simple subtraction is done, eg., 



Ho: pi- y 2=0 vs Ha: yl- y 2=d 



where d is some specified difference. For log transformed data this 

 difference is the ratio of the geometric means. Therefore a difference, 

 d, in the means of log transformed data is equivalent to a ratio, R, of 

 the geometric means of the untransf ormed data (d=log (R)). To test 

 against an alternate hypothesis that the geometric means differ by 30%, 

 we set R=1.30 and thus d=0.114. 



To estimate the variance of the transformed data, the following 

 analysis of variance model was used: 



L. . = u + T, + S. + e. . (6) 



where u = mean 



T = effect of the i sampling time or data 

 S = effect of the j th station 

 e = error term 



2 

 This error term was used as an estimate of a in determining the sample 



size, N, required to detect a specified difference, d, at various a and 



levels (Snedecor and Cochran, 1967) : 



N = (Z Q + Zg) 2 * 2o 2 / d 2 (7) 



where Z = normal deviate such that the area from -°° to Z 

 a is (1 - a) 

 Zg = normal deviate such that the area from -°° to Z 

 is (1 - 3) 

 d = lo SlO R 

 a = error variance estimated from the ANOVA 



The cost effectiveness of the seine program was evaluated following 



the method outlined by TASC (1979). The objective was to minimize the 



variance of the seine program within the current cost. The AOV model 



used was the same as equation (6). The model used to estimate yearly 



cost was given by: 



43 



