How it works
Four steps, in the order the page asks for them.
What the calculator provides
Every design the calculator covers
Pick a test type in the sidebar and the design panel changes to the inputs that test needs. The effect size is always reported alongside the required sample size.
Means
- Means (1-sample)
- One-sample t-test against a reference value, reported with Cohen's d.
- Means (2-sample)
- Two-sample t-test with Cohen's d and an allocation ratio when the groups are not the same size.
Proportions
- Proportion (1-sample)
- A single expected proportion tested against a null value.
- Proportions (2-sample)
- Two independent proportions, entered as the expected p₁ and the comparison p₂.
Association and models
- Correlation
- An expected correlation against a null value, planned through Fisher's z transformation.
- ANOVA
- One-way ANOVA for k groups using the effect size f, planned for the omnibus test.
- Regression
- Linear regression from the number of predictors and the R² you expect to explain.
What the numbers rest on
Worth knowing before you quote a sample size in a protocol.
- Power defaults to 0.80 and is the target the calculator solves for; the power actually achieved at the returned whole-number sample size is reported separately, because rounding up usually overshoots slightly.
- One-sample means and one-way ANOVA are computed through non-central distributions rather than a normal approximation, which matters most at small sample sizes.
- The power curve plots required sample size against power across the range, so you can see how much a study costs to move from 0.80 to 0.90 before committing.
- Results are planning figures for a single primary comparison. They carry no adjustment for dropout, clustering, interim analyses or multiple endpoints.
Common questions
What does this calculator solve for?
The sample size required to reach a target power, given your significance level, hypothesis direction and expected effect. It also reports the power actually achieved at that sample size and draws the sample-size-against-power curve.
Which tests are supported?
One-sample and two-sample means, one-sample and two-sample proportions, correlation, one-way ANOVA and linear regression.
Can I plan an unequal allocation between two groups?
Yes, for two-sample means. An allocation field sets the ratio of the second group to the first, and the required size of each group is returned.
Does it handle non-inferiority, dropout or clustered designs?
No. It plans conventional superiority tests of a single comparison, with no dropout inflation, no intracluster correlation and no equivalence or non-inferiority margins.
Do I need Excel or an install to use it?
No. It runs in the browser. It is also reachable from inside the Statistico Excel add-in, where a design can be sent straight from an analysis into the calculator.