What is statistical power?

By Dr. Rafiq Muhammad, MD, PhD · Updated June 2026

Statistical power is the probability that your study will detect a real effect of a given size, if that effect truly exists. Power of 0.80 means an 80% chance of finding the effect — and a 20% chance of missing it (a Type II error). Power is the single best defence against the most demoralising research outcome: a true effect that your study was simply too small to see.

The four quantities that interlock

Power doesn’t stand alone. It’s locked together with three others, and fixing any three determines the fourth:

• Power — your chance of detecting the effect (aim for ≥ 0.80).
• Sample size — how many participants you recruit.
• Effect size — how big the effect you want to detect is.
• Significance level (alpha) — your false-positive risk, usually 0.05.

Bigger samples, larger effects, and a more lenient alpha all raise power. That’s why a tiny study chasing a small effect is almost guaranteed to be underpowered.

Why 80% — and why you decide it before you start

80% is the conventional minimum (90% is stronger). The crucial point is when you compute it: a proper a priori power analysis happens before data collection — you fix power, your target effect size, and alpha, then solve for the sample size to recruit. Computing power after the fact from your observed effect (“post-hoc” or “observed” power) is circular and tells you nothing useful — reviewers will flag it.

How to run a power analysis

You need a realistic target effect size — ideally from a prior study or a pilot, not a hopeful guess (over-estimating the effect is the most common route to an underpowered study). Then plug power, effect, and alpha into a calculator to get your sample size; or enter a sample size you can realistically recruit to see the power it buys you.