Since the aim was to interpret CCA using biplot representations
(Gabriel, 1971) of either weight or correlation structure matrices,
we determined how many dimensions were necessary to explain
the relationships between both sets of latent variables. According
to this, the first and second two canonical variates were the only
significant ones (p < 0.001) (Table 4) besides having the highest singular
values among all possible correlations (ri > 0.886 for all FMP)