Wide Linear discriminant analysis is performed as follows:
• The data are standardized by subtracting group means and dividing by pooled standard deviations.
• The singular value decomposition is used to obtain a principal component transformation matrix from the set of singular vectors.
• The number of components retained represents a minimum of 0.9999 of the sum of the squared singular values.
• A linear discriminant analysis is performed on the transformed data, where the data are not shifted by group means. This is a fast calculation because the pooled-within covariance matrix is diagonal.