7 5 



'. i' \' 



assemblage and demonstrated stronger correlations with TSI(AVG) (r 

 :^ ■ = 0.783, p < 0.001, n = 29) and pH (r = 0.885, p < 0.001, n = 29) than 

 r • the procedure with untransformed concentration data. None of the 

 correlation coefficients between principal components and 

 macrophyte variables were significant in this procedure, even when 

 the effects of TSI(AVG) and pH were partialled out. 



Principal components analysis using annual diatom accumulation 

 rates produced a first principal component that explained 21.4% of 

 the variance and was again significantly correlated with TSI(AVG) (r 

 = 0.460, p = 0.012, n = 29), pH (r = 0.464, p = 0.011, n = 29), and 

 specific conductance (r = 0.501, p = 0.006, n = 29). The only 

 significant correlation coefficient involving a macrophyte variable 

 was with the fourth principal component that was significantly 

 correlated with percent-area coverage (r = -0.448, p = 0.036. n = 22) 

 and more strongly correlated with TSI(AVG) (r = 0.501, p = 0.006, n = 

 29). A predictive model using eigenvectors of this principal 

 component would not be useful for predicting percent-area coverage 

 because it would be confounded by TSI(AVG). 



Principal components based on diatom accumulation rates with 

 the effects of TSI(AVG) and pH partialled out produced a single 

 significant correlation with a macrophyte variable. The sixth 

 principal component, which accounted for 5.5% of the variance in the 

 diatom assemblages, had a significant negative correlation with 

 floating-leaved biomass (r = -0.574, p = 0.005, n = 22). A predictive 

 model using the eigenvectors for this principal component would 

 explain 33.0% of the variance in floating-leaved biomass but could 



