ABSTRACT
The traditional null hypothesis is that treatments, interventions, etc., have no effect; in Chapters 1 and 2, we use the term nil hypothesis to describe this particular version of H0. The nil hypothesis is so common and so widely used that most researchers assume that the hypothesis that treatments have no effect, or that the correlation between two variables is zero, is the null hypothesis. This is wrong. The null hypothesis is simply the specific hypotheses that is being tested (and that might be nullified by the data), and there are an infinite number of null hypotheses researchers might test. One researcher comparing two treatments might test the hypothesis that there is no difference between the mean scores of people who receive different treatments. A different researcher might test the hypothesis that one treatment yields scores that are, on average, 5 points higher than those obtained using another treatment. Yet another researcher might test the hypothesis that treatments have a very large effect, accounting for at least 25% of the variance in outcomes. These are all null hypotheses. Knowing that there are so many null hypotheses that might be tested, it is useful to understand why one special form (i.e., the nil hypothesis) is the one that actually is tested in most statistical analyses.
