What are the disadvantanges of canonical correlation analysis?

Wenlong

Hi, all

I am looking into canonical correlation analysis recently, and trying to dig deeper into it.

So far, I know the advantages of canonical correlation analysis are that it capable of revealing correlation between two sets of multi-dimensional variables. It also reduces the risk of committing Type I error.

However, I don't know about its disadvantages, though I believe there must be some.

The following link is the slides of a conference paper: http://www.math.univ-toulouse.fr/~ig...rance_Vzla.pdf

The authors claimed that the disadvantages of CCA are: 1. Much information get lost; 2. Poor interpretation of the results. Nontheless, there are no explainations about them.

May I know what are the disadvantages of canonical correlation analysis? And what nature of CCA caused these disadvantages?

Many thanks in advance for your kindly help.

All bests
Wenlong