
240 ◾ Software Metrics
alternative hypotheses, and rejection of H
0
simply means that more exper-
imentation is needed to determine which alternative hypothesis is the best
explanation of the observed behavior.
We emphasize that statistical analysis is directed only at whether we
can reject the null hypothesis. In this sense, our data can refute the alter-
native hypothesis in light of empirical evidence (i.e., the data support the
null hypothesis because there is no compelling evidence to reject it), but
we can never prove it. In many sciences, a large body of empirical data is
amassed, wherein each case rejects the same null hypothesis; t