MULTIPLE REGRESSION 87
to be very useful. Thus the decision about how useful a model is, at least
once it has been found to be statistically significant, should be theory
based and not statistically based. This requires knowledge of the
field under study rather than statistics. This is the main reason that
statisticians usually need help from knowledgeable outsiders when
developing models.
Example
We will now look at an example involving celebrities. Although this is
clearly a nonbusiness example (unless, of course, you are in the business
of making movies), the procedures and considerations are exactly the
same as performing a marketing analysis to try to understand why your
product is (or is not) popular.
Forbes collected ...