In the canonical correlations program, 24 variables
were grouped into sets. The sets were then compared and
significant relationships were defined. The second program
was a full multivariate regression model which explained
the significance of each of the measured variables in
influencing alligatorweed growtn.
In canonical correlations, a set of independent vari-
ables may be compared with a set of dependent variables to
find the linear combination of variables in each set which
when correlated is maximum. The resultant variable is
known as a canonical variate. If some linear relationship
between the sets of variables still remains unaccounted for
by the first set of canonical variates, the process of find-
ing new linear combinations that would best account for the
resideual relationships between the sets can be continued.
This process can go on until there are no significant
linear associations left. Each canonical variate is
orthogonal to (or unrelated to) other canonical variates.
The chisquare test is used to evaluate variates for signi-
ficance.
In canonical correlations, variables in one set may
be combined to predict maximally the variations of the
variables in the other set. This process was utilized to
compare data from lakes and streams, and to predict the
number of samples that should occur in using different sets,
lakes or streams (See Morrison 1967) for detailed discussion
of canonical correlations and tests of significance). All