Suppose that is represented by a cloud of points (variables) in (considering each column). How can this cloud be projected into a lower dimensional space? We start as before with one dimension. In other words, we have to find a straight line , which is defined by the unit vector , and which gives the best fit of the initial cloud of points.
Algebraically, this is the same problem as above (replace by and follow Section 8.2): the representation of the -th variable is obtained by the projection of the corresponding point onto the straight line or the direction . Hence we have to find such that is maximized, or equivalently, we have to find the unit vector which maximizes . The solution is given by Theorem 2.5.
The coordinates of the variables on are given by
, the first factorial axis.
The variables are now represented by a linear combination of the
original individuals
, whose coefficients are
given by the vector , i.e., for
(8.5) |
The representation of the variables in a subspace of dimension is done in the same manner as for the individuals above. The best subspace is generated by the orthonormal eigenvectors of associated with the eigenvalues . The coordinates of the variables on the -th factorial axis are given by the factorial variables . Each factorial variable is a linear combination of the original individuals whose coefficients are given by the elements of the -th vector . The representation in a subspace of dimension is depicted in Figure 8.5.