Principal components analysis

finds linear combinations of a set of variates that successively maximize the variation contained within them. So the first combination (or principal component) gives the single direction containing the most variation, the first two components define the plane containing the most variation, and so on. (See PCP.)