Page 142 - Jolliffe I. Principal Component Analysis
P. 142

6


                              Choosing a Subset of Principal
                              Components or Variables


















                              In this chapter two separate, but related, topics are considered, both of
                              which are concerned with choosing a subset of variables. In the first section,
                              the choice to be examined is how many PCs adequately account for the
                              total variation in x. The major objective in many applications of PCA is
                              to replace the p elements of x by a much smaller number m of PCs, which
                              nevertheless discard very little information. It is crucial to know how small
                              m can be taken without serious information loss. Various rules, many ad
                              hoc, have been proposed for determining a suitable value of m, and these
                              are discussed in Section 6.1. Examples of their use are given in Section 6.2.
                                Using m PCs instead of p variables considerably reduces the dimension-
                              ality of the problem when m   p, but usually the values of all p variables
                              are still needed in order to calculate the PCs, as each PC is likely to be
                              a function of all p variables. It might be preferable if, instead of using m
                              PCs we could use m, or perhaps slightly more, of the original variables,
                              to account for most of the variation in x. The question arises of how to
                              compare the information contained in a subset of variables with that in
                              the full data set. Different answers to this question lead to different criteria
                              and different algorithms for choosing the subset. In Section 6.3 we concen-
                              trate on methods that either use PCA to choose the variables or aim to
                              reproduce the PCs in the full data set with a subset of variables, though
                              other variable selection techniques are also mentioned briefly. Section 6.4
                              gives two examples of the use of variable selection methods.
                                All of the variable selection methods described in the present chapter
                              are appropriate when the objective is to describe variation within x as
                              well as possible. Variable selection when x is a set of regressor variables
   137   138   139   140   141   142   143   144   145   146   147