How many participants do I need?
Don’t have d? Enter a raw difference and SD instead:
Get the free Research Design toolkit
Sampling, power, and study-design templates & checklists from Research Design Simplified. We’ll email you the download link.
How to calculate sample size
Every sample size comes from four ingredients: the effect you want to detect (or the margin of error you can accept), your confidence level (how sure you want to be — commonly 95%), and, for comparison studies, your statistical power (the chance of detecting a real effect — commonly 80%). Bigger confidence and power, or a smaller effect, all push the number up.
- Comparing two means: driven by Cohen’s d (or a mean difference ÷ SD), confidence, and power.
- Comparing two proportions: driven by the two expected rates, confidence, and power.
- Estimating one mean or proportion: driven by the SD (or expected proportion) and the margin of error you’ll accept.
- Detecting a correlation: driven by the size of r you want to detect.
- Always inflate for dropout: if you expect attrition, recruit more so the analysable sample still meets the target.
This calculator covers the common two-sided designs and is a planning aid, not a substitute for a formal power analysis on complex designs (clustering, repeated measures, survival, non-inferiority). For those, confirm with a statistician.
Frequently asked questions
How do I calculate sample size?
Choose your design, set the effect you want to detect (or the margin of error you can accept), your confidence level (commonly 95%), and — for comparisons — your statistical power (commonly 80%). The calculator converts those into the number of participants you need.
What statistical power should I use?
Power is the chance of detecting a real effect of the size you specify. 80% is the conventional minimum; 90% is stronger but needs a larger sample. Power applies to comparison studies, not simple estimation.
What effect size should I enter?
The smallest difference that would be meaningful in your field — ideally from prior studies or a pilot. Inflating the effect to shrink the sample is the most common route to an underpowered study.
Does this account for dropout?
No — it returns the analysable sample size. If you expect d% attrition, recruit n ÷ (1 − d/100).
Does it store anything?
No. The calculation runs entirely in your browser; nothing is uploaded or saved.