ABSTRACT
The power of a statistical test is the probability that the study reject the null hypothesis being tested, given that this null hypothesis is in fact wrong. As noted throughout, the traditional null hypothesis that treatments have no effect whatsoever (or that the correlation between two variables is precisely zero, or any other hypothesis of “no effect”) is very often wrong, and in this context the statistical power of a test is essentially the probability that the test will lead to the correct conclusion. When testing the traditional null hypothesis, it is obvious that power should always be as high as possible. When testing a minimum-effect hypothesis (e.g., that the effect of treatments is negligibly small, but not necessarily precisely zero), the implications of varying levels of statistical power are potentially more complex, and a wider range of issues needs to be considered in determining how to use and interpret statistical power analysis.
