The importance of statistical power

As currently practiced in the social and health sciences, inferential statistics rest solidly upon two pillars: statistical significance and statistical power. The two concepts, both of which are expressed in terms of probabilities (i.e., how likely events are to occur), are so integrally related to one another that it is almost impossible to consider them separately.

Statistical significance, the first pillar, is a probability level generated as a byproduct of the statistical analytic process. It is computed after a study is completed and its data are collected. It is used to estimate how probable the study's obtained difference or relationship (which is called its effect size) would be to occur by chance alone. Based in large part upon Sir Ronald Fisher's (1935) recommendations, this probability level is often interpreted as an absolute standard. If it is 0.05 or below, the results are said to be statistically significant and the researcher has, by definition, supported his/her hypothesis. If it is 0.06 or above, then statistical significance is not obtained and the research hypothesis is not supported.

Statistical power, on the other hand, is computed before a study's final data are collected. It involves a two-step process: (a) hypothesizing the effect size which is most likely to occur based upon the study's theoretical and empirical context and (b) estimating how probable the study's results are to result in statistical significance if this hypothesis is correct.