Statistical power explained, with a worked example
Last reviewed 16 October 2026
Power is the probability that a study detects a real effect of the size you specified. It depends on the sample size, the size of the effect, how variable the outcome is and the significance level.
Open the power calculatorWhat 80% power means
With 80% power a study misses a true effect of the stated size one time in five. That is why 80% is treated as a minimum, and why definitive trials often aim for 90%.
A worked example
Take infection falling from 30% to 20%, with a two-sided significance level of 0.05. With 294 per arm the power is 80%. With only 150 per arm it falls to 52%: about one chance in two of finding a real effect that exists.
What you can detect with fixed numbers
Often the sample is fixed by budget or by who is available. Then ask what the study can detect. With 100 per group and a standard deviation of 15, the smallest difference detectable with 80% power is about 5.9.
Do not calculate power after the study
Power calculated from the observed effect adds nothing to the p-value. Report the confidence interval instead: it shows which effects the data rule out.
Underpowered studies
A study with low power is likely to miss real effects, and the significant results it does find tend to overstate them. Ethics committees increasingly ask whether a study is large enough to be worth the burden on participants.
Common questions
- What is a good level of power?
- At least 80%, and 90% for a definitive trial.
- How do I increase power?
- Increase the sample, target a larger difference, reduce variability (a more precise measurement, a paired design or adjustment for baseline), or accept a higher significance level, which is rarely appropriate.
- Is power the same as the p-value?
- No. Power is planned before the study from assumed values. A p-value is calculated from the data after it.
Keep reading
General methods guidance, not a substitute for advice from a statistician on your specific study.