The whitening transformation is a linear preprocessing method which allows correlation among variables. The transformation is done by using eigenvectors from covariance matrix in a way that is illustrated rougly in the figure below. Data frames can only contain floating point vectors. This transformation is useful when data contains variables which differ greatly in magnitude.
Example (ex4.12): Load some arbitrary data and make the whitening transformation.
# just load some data in NDA> load sin.dat # now make transformation NDA> whitening -d sin -dout whitesin ...