Power analysis is a statistical technique that is used to determine the sample size needed for a study in order to detect a difference of a specified size with a specified level of confidence. It helps to ensure the reproducibility of a study by allowing other researchers to see exactly how the sample size (e.g. number of participants) was determined and to evaluate whether it was appropriate for the research question being asked.
Another benefit of power analysis is that it can help to reduce the likelihood of false positives and false negatives in a study. A false positive is when a study finds a statistically significant difference between two groups when in fact there is no real difference. A false negative is when a study fails to find a statistically significant difference between two groups when in fact there is a real difference. Power analysis can help to reduce the likelihood of these errors by ensuring that the sample size is large enough to detect the effect of interest.
Underpowered studies often lead to reproducibility problems. A power analysis can prevent this from the outset by determining your statistical power. There are four pillars that you should keep in mind when performing such an analysis: 1) effect size, 2) sample size, 3) significance and 4) statistical power. For a more detailed explanation of the concept of power analysis, please see here.
- When Power Analyses Based on Pilot Data are Biased: Inaccurate Effect Size Estimators and Follow-up Bias (Albers & Lakens, 2018)
- Using Anchor-Based Methods to Determine the Smallest Effect Size of Interest (Anvari & Lakens, 2019)
- Correcting for Bias in Psychology: A Comparison of Meta-Analytic Methods (Carter et al., 2019)
- Safeguard Power as a Protection Against Imprecise Power Estimates (Perugini et al., 2014)
- A tutorial on Bayes Factor Design Analysis using an informed prior (Stefan et al., 2019)
- G*Power (a tool to compute power analyses for many different statistical tests)
- R package pwr (statistical power analysis in R)
- R package BFDA (for Bayesian sample size planning)