Statistical Power Analysis

Preprint of Chapter appearing as: Aberson, C. L. (2015). Statistical power analysis. In R. A. Scott & S. M. Kosslyn (Eds.) Emerging trends in the behavioral and social sciences. Hoboken, NJ: Wiley.Statistical power refers to the probability of rejecting a false null hypothesis (i.e., finding what the researcher wants to find). Power analysis allows researchers to determine adequate sample size for designing studies with an optimal probability for rejecting false null hypotheses. When conducted correctly, power analysis helps researchers make informed decisions about sample size selection. Statistical power analysis most commonly involves specifying statistic test criteria (Type I error rate), desired level of power, and the effect size expected in the population. This article outlines the basic concepts relevant to statistical power, factors that influence power, how to establish the different parameters for power analysis, and determination and interpretation of the effect size estimates for power. I also address innovative work such as the continued development of software resources for power analysis and protocols for designing for precision of confidence intervals (a.k.a., accuracy in parameter estimation). Finally, I outline understudied areas such as power analysis for designs with multiple predictors, reporting and interpreting power analyses in published work, designing for meaningfully sized effects, and power to detect multiple effects in the same study.