Describe the issue linked to the documentation

The user guide documentation for EmpiricalCovariance currently states:

More precisely, if assume_centered=False, then the test set is supposed to have the same mean vector as the training set. If not, both should be centered by the user, and assume_centered=True should be used.

It doesn't make sense, however, that assume_centered=False would require data to be centered. Likewise, it would seem that the user would need to center the data OR use assume_centered=True -- not both.

Additionally, it doesn't seem like there are separate training and testing data for this.

Suggest a potential alternative/fix

I think it should read:

More precisely, if assume_centered=True, then the data set's mean vector should be zero. If not, the data should be centered by the user, or assume_centered=False should be used.