A k-means-type algorithm is proposed for efficiently clustering data constrained to lie on the surface of a p-dimensional unit sphere, or data that are mean-zero-unit-variance standardized ...