where: 



ng = 3 levels of seasonal effort (3 days) 



Op = 2 levels of diurnal effort (day-night) 



nj^ = 2 levels of net effect (2 nets) 



r =24 replicates (2 x 12) from 2 tow directions and 12 hours within 

 each of the 2 levels of the diurnal effect 



m =13 missing samples 



Due to uniform environmental conditions over the three sampling days the SD 

 (seasonal - diurnal) interaction is not considered of interest. The second order 

 interaction, SND (seasonal - diurnal - nets) is considered unrealistic and dropped 

 from consideration. The general linear model then reduces to: 



Yjjki = /x + Si + Dj + Nk + SNjk + DNjk + Ejjkl 



The missing terms SDj: and SNDjjk are pooled with the random error term E jjki 

 According to the above model, the partitioning of the available degrees of freedom 

 (N 1 = 274) for testing resuhs in the univariate analyses of variance Table shown as 

 Table 22.2. The high sensitivity of the test is guaranteed by the large number of 

 degrees of freedom available for the error term ( W = 267). This analysis of variance 

 model may be applied to several response variables, including eggs, larvae and 

 zooplankton. 



In addition to linearizing the periodic responses to diurnal and tidal cycles, their 

 variance components may be removed prior to testing for these responses known to 

 be linearly correlated with time. The approach will involve spectral analysis and 

 periodic regression. 



343 



