Sample size calculator
Work out how many participants you need to compare two groups or detect a correlation with adequate statistical power, or to estimate a proportion or mean to a target margin of error. Free, no sign-up; everything runs in your browser.
Uses standard normal-approximation formulas (Cohen, 1988; survey CI). For comparison tests, add ~10–15% for expected dropout, and consider exact / simulation-based power for small samples or complex designs. Everything runs in your browser.
Designing the study?
Plan the sample here, run the review with Folio's meta-analysis and PRISMA tools, then write it up with citation-checked drafting.
Why plan your sample size?
An underpowered study risks missing a real effect; an oversized one wastes resources and can raise ethical concerns. A sample-size calculation, done before you collect data, ties your target effect size, significance level, and desired power to the number of participants you actually need, and it's expected in pre-registrations, grant applications, and ethics submissions.
This calculator pairs with Folio's free meta-analysis and PRISMA tools. Plan the study here, then analyze and write it up in one workspace.
Frequently asked
Is this sample size calculator free?
Yes, completely free, with no sign-up. All calculations run in your browser; nothing is uploaded.
What does statistical power mean?
Power (1 − β) is the probability of detecting an effect that truly exists. 80% is the conventional minimum; 90% is common when a missed effect is costly. Higher power requires a larger sample.
What is Cohen’s d?
A standardized effect size for a difference in means: the difference divided by the standard deviation. Conventional benchmarks are 0.2 (small), 0.5 (medium), and 0.8 (large), but use an estimate from prior work or a pilot when you can.
How do I find the sample size for a correlation?
Choose “Detect a correlation” and enter the correlation (r) you expect, plus your significance level and power. It uses Fisher’s z transformation (Cohen, 1988): detecting r = 0.30 at 80% power and α = .05 needs about 85 participants, while a smaller r = 0.10 needs around 783. Weaker correlations require much larger samples.
How is the survey sample size calculated?
For estimating a proportion, it uses n = z²·p·(1−p) / margin², with an optional finite-population correction when you know the population size. With p unknown, 0.5 gives the most conservative (largest) sample.
How accurate are the numbers?
These are standard normal-approximation formulas used by most calculators and are excellent for planning. For small samples, sequential designs, clustering, or complex models, confirm with exact or simulation-based power (e.g. G*Power or a stats package). Add a margin for expected dropout.