Means rotation

For each group pair, computes a normalized mean-difference axis on the progressively-deflated data and finishes with lq_orthogonal_svd. The input is column-centered first, matching rENA's scale(data, scale=F, center=T) at the top of ena.rotate.by.mean.

lq_means_rotation(points, group_pairs)

Arguments

points

Numeric matrix (n_units x n_dims)

group_pairs

List of length k; each element is list(a, b) where a and b are 0-based integer index vectors

Value

List with rotation, eigenvalues, column_names

Details

Each element of group_pairs is a length-2 list list(a, b) of 0-based row indices into points.