Explained covariance?


This may reflect my lack of understanding of PLS, but the explained_variance attribute gives the percent variance explained by each component for X and for Y. Is there a way to find the percent covariance between X and Y explained by each component?


1 Like

hi @user2842,

The easiest way would be to simply calculate the sample covariance(object$variates$X[,1], object$variates$Y[,1]) for example for component 1,=.

To get the total covariance explained: Rd(X, Y) = Rd(X, object$variates$Y), and Rd(Y,X) = Rd(Y, object$variates$X) but that assumes you have calculated all possible dimensions. At this stage I can’t really comment on whether this is possible. Let me know if you find a better answer.


what function is Rd?

Rd is not a specific function within the mixOmics package. The formula in the above excerpt describes a measure (‚Äúredundancy‚ÄĚ) of the explained variance of a given component.

If you are wanting to examine the source code used to calculate the explained variance, look at the explained_variance() function - found here.

Dear all,

I am very interesting in the concept of ‚Äėexplained_covariance‚Äô, so please let me know if there is any update on that topic!
Moreover, I don‚Äôt really understand why ‚Äėexplained_variance‚Äô seems quite simple to compute, but not ‚Äėexplained_covariance‚Äô‚Ķ :confused: