Free research tools

Sample size, effect size and test chooser calculators

Quick, transparent calculators for the questions supervisors and ethics committees ask first. Each one shows the formula, so you can cite it in your methods chapter. Everything runs in your browser; nothing you type is sent or saved.

Survey sample size (proportion)

For estimating a percentage, for example "what share of nurses use AI tools", within a chosen margin of error.

Formula

n₀ = z² × p(1 − p) / e². With a known population N: n = n₀ / (1 + (n₀ − 1) / N). Invitations = n ÷ response rate. Using p = 50% gives the most cautious (largest) sample.

Sample size to compare two groups (power analysis)

For a study comparing two independent groups on a continuous outcome, such as an intervention versus a control group.

Formula

n per group = 2 × (z₁₋α/₂ + z₁₋β)² / d², rounded up, plus one per group as a small correction for using the t-distribution. Then inflate for drop-out: n ÷ (1 − drop-out). Don't know d? Use the smallest difference that would matter in practice divided by the expected standard deviation, or an estimate from a pilot or a similar published study.

Effect size from two group means (Cohen's d and Hedges' g)

Turn group means and standard deviations, from your data or from a published paper, into a standardised effect size.

Formula

Pooled SD = √[((n₁ − 1)s₁² + (n₂ − 1)s₂²) / (n₁ + n₂ − 2)]. d = (M₁ − M₂) / pooled SD. Hedges' g = d × (1 − 3 / (4(n₁ + n₂) − 9)), which corrects the small-sample bias and is the usual choice in meta-analysis.

Which statistical test should I use?

Answer three questions to get the usual test for your design, with where to find it in SPSS and the R command.

This covers the common designs. Clustered data, many outcomes or complex sampling need a tailored analysis plan.

Frequently asked questions

How do I calculate sample size for a survey?

Use n0 = z² × p(1 − p) / e², where z is 1.96 for 95% confidence, p is the expected proportion (0.5 if unknown, which gives the largest sample) and e is the margin of error as a decimal. If the whole population is small, correct it with n = n0 / (1 + (n0 − 1) / N). Then divide by your expected response rate.

What sample size do I need to compare two groups?

For two independent groups compared on a continuous outcome, each group needs about 2 × (z for alpha + z for power)² / d², where d is the smallest effect size worth detecting. With alpha 0.05, 80% power and a medium effect (d = 0.5) that is about 63–64 per group. Smaller effects need far larger samples.

What is a good effect size?

Cohen's rough benchmarks for d are 0.2 small, 0.5 medium and 0.8 large, but the meaningful size depends on your field and outcome. Report the effect size with a confidence interval rather than relying on the labels.

Should I use a t-test or Mann–Whitney U?

Use the independent-samples t-test (preferably Welch's version) for a roughly normal continuous outcome or a reasonably large sample. Use Mann–Whitney U for ordinal data, such as Likert items, or for small samples with clearly skewed data.

Are these calculators enough for an ethics or grant application?

They are a sound first estimate and show the formula you can cite. For complex designs (clusters, repeated measures, regression with many predictors, survival outcomes) run a formal power analysis in G*Power, R (pwr) or a statistician's software, and state your assumptions.

Want the analysis checked?

Rahmat can review your sample size justification, analysis plan or results chapter, and explain every choice so you can defend it in your viva or to reviewers.

Work with Rahmat

Send your stage, field, word count and deadline. You get a written scope and a fixed price before anything starts.

  • Free first review of your brief
  • Milestones, revisions within scope
  • Confidential; NDA on request