Canonical Correspondence Analysis with variation partitioning: some comments and an application