The Knoll Hypothesis and the Alternative Hypothesis

In a factorial kind of way, even for something as simple as assessing a mean difference, you can make a part of your model all kinds of different aspects. Even in the test of mean differences, that the mean differences is five points between these two groups, that could be a competing model. We're very cavalier with zero, but we don't go out on a limb. And what we talked about last week with roy is that we are generally very cowardly when it comes to specifying alternatives as being free perimeters rather than fixed perometers.