
Determine from the selected CCA and Correlation objects the fraction of the variance explained by the selected canonical variate range.
1. In general the variance fractions for a particular canonical variate in the dependent and in the independent set are not the same.
2. In general, the variance fractions for all canonical variates do not sum to 1. (The technical reason is that for canonical correlation analysis in general the eigenvectors are not orthogonal, i.e. they overlap and therefore, necessarily, also the variance fractions overlap.)
The formulas can be found on page 170 of Cooley & Lohnes (1971).
For example, the fraction of the variance explained by the i^{th} canonical variable in the dependent set is:
fractionVariance = ((y_{i}′ R_{yy}′ R_{yy} y_{i}) / (y_{i}′ R_{yy} y_{i})) / n_{y}, 
where y_{i} is the eigenvector for dependent canonical variable i and R_{yy} is the correlation matrix for the n_{y} variables in the dependent set.
© djmw 20181112