40 



Canonical correspondence analysis (CCA) was performed on the 

 percentage data for 47 diatom taxonomic groups using the canonical 

 community ordination (CANOCO) statistical package developed by ter 

 Braak (1987). In the first set of computations, the diatom groups 

 were ordinated in an axis constrained by the environmental variable 

 percent-volume infestation. In a second set of computations, 47 

 diatom taxa were ordinated into an axis constrained by percent-area 

 coverage. 



Multivariate models that predict percent-volume infestation, 

 percent-area coverage and biomass for submerged, emergent and 

 floating-leaved plants were derived using the maximum R2 

 improvement method of the SAS STEPWISE procedure (SAS Inst., Inc. 

 1985). I reduced the number of independent diatom variables in 

 each stepwise regression to 20 or less, a recommended maximum for 

 this procedure (SAS Inst., Inc. 1985), after examining plots of diatom 

 taxonomic groups versus macrophyte variables. Regressions were 

 performed for each macrophyte variable using percentage data for 

 the diatom groups. The regressions for percent-area coverage and 

 percent-volume infestation were repeated using concentration and 

 accumulation rate data for the diatom groups. 



I selected the best model in each STEPWISE regression 

 procedure by plotting Mallows' Cp statistic versus the number of 

 variables in each model (p) and selecting the model in which Cp was 

 approximately equal to p (Daniel and Wood 1971). Adjusted R^s, 

 which show the coefficient of determination after removing the 

 inflating effect of dependent variables, were calculated using the SAS 

 RSQUARE procedure (SAS Inst., Inc. 1985). 



