were significant at the .01 probability level: PH, alkalinity,
iron, turbidity, chlorides, phosphates, nitrates, hardness,
and water depth. The result was an overall accuracy of 99
percent. Ninety-three of the 96 lake samples were reclassi-
fied correctly with only three appearing as stream samples
(97% accuracy). The stream samples were reclassified with
100 percent accuracy (Table 13).
The author is cognizant of the fact that with the
small number of sample sites these comparisons may not
represent a true picture of lakes and streams in general;
they are, however, valid for this study.
Analysis of variance of Plant Characteristics
Two analysis of variance tables were constructed for
each of the variables representing plant characteristics;
one which showed the statistical significance of the
dependent variable with all independent variables included
in the model, and the other an analysis of the dependent
variable with only the independent variables representing
the insect populations included in the model.
In each analysis, the sums of squares was broken down
using partial sums of squares to show how each of the
independent variables contributed to the variation of the
dependent variable. In each case the F test was employed
to determine statistical significance.
As shown in table 18 plant height was significant at
greater than the .01 probability level in both analysis.
The partial sums of squares indicate that iron, turbidity,