Plan the right number of respondents before you collect data — or check the power of a study you've already run.
What it does
This tool answers two of the most common planning questions in survey and experimental research: "How many respondents do I need?" and "Was my study big enough to detect the effect I was looking for?" It covers five common scenarios — estimating a mean, estimating a proportion, comparing two means, comparing two proportions, and calculating the achieved power of a two-group comparison you've already conducted.
Who it's for
Built for students, thesis researchers, and academicians who need to justify a sample size in a research proposal or methodology chapter, or who want to check — before submitting a survey to a review committee, or before writing up results — whether their planned or actual sample size was adequate.
ЁЯУЦ Theory & Formulas — Which One Do I Need?
Every formula below answers a different question. Pick the scenario that matches your research design, then use the matching panel in the tool below. All formulas here use the normal (Z) approximation, the standard method taught in research-methods courses and used in most published sample-size tables (Cochran, 1977; Cohen, 1988; see also Chow, Shao & Wang, 2008, and Julious, 2009, for the wider methodology). They are accurate for typical planning purposes but are approximations — for very small planned samples, dedicated software (e.g. G*Power) that iterates on the exact non-central t distribution will give a slightly more precise number.
1. Estimating a Population Mean
Use this when: you're running a descriptive/survey study and want to estimate an average (e.g. average yield, average score) within a chosen margin of error.
- n₀ — required sample size
- Z╬▒/2 — critical value for your confidence level (e.g. 1.96 for 95%)
- ╧Г — expected population standard deviation (from a pilot study or similar past research)
- E — the margin of error you're willing to accept, in the same units as ╧Г
If you know your total population size N (e.g. all farmers in a district), apply the finite population correction: n = n₀ / (1 + (n₀−1)/N) — this shrinks the requirement when the population itself is small.
2. Estimating a Population Proportion
Use this when: you're estimating a percentage (e.g. % of farmers using a practice, % of students who agree with a statement).
- p — your best estimate of the true proportion (0 to 1). If you have no prior estimate, use p = 0.5 — it maximizes p(1−p) and gives the most conservative (largest, safest) sample size.
- E — margin of error as a proportion (e.g. 0.05 for ±5 percentage points)
The same finite population correction as above applies if N is known.
3. Comparing Two Independent Means
Use this when: you're designing an experiment or comparison (e.g. treatment vs. control) and want enough participants per group to detect a real difference, if one exists — this is a power analysis, not just an estimation problem.
- Z╬▓ — critical value for your desired power (e.g. 0.84 for 80% power)
- d — Cohen's effect size: d = (╬╝₁ − ╬╝₂) / ╧Г, the expected difference in means divided by the (assumed common) standard deviation
Choosing d — Cohen's conventional benchmarks: d = 0.2 (small effect), 0.5 (medium), 0.8 (large). Use a pilot study's numbers if you have them; otherwise "medium" is a common default assumption.
4. Comparing Two Independent Proportions
Use this when: comparing a success rate, adoption rate, or agreement rate between two groups.
- p₁, p₂ — the two proportions you expect to compare (your best estimates, e.g. from a pilot or literature)
5. Approximate Achieved (Post-Hoc) Power for a Two-Group Comparison
Use this when: the study is already done — you know the sample size you actually used and want an estimate of how much power it gave you to detect a given effect size.
- ╬ж — the standard normal cumulative distribution function
- n — the sample size per group you actually used
Conventionally, power ≥ 0.80 (an 80% chance of detecting the effect if it's real) is considered acceptable for most social-science and agricultural research.
Note: This calculator uses the standard normal approximation to estimate statistical power. Dedicated software such as G*Power uses the exact non-central t distribution and may produce slightly different values, particularly for small sample sizes.
Understanding ╬▒ (alpha), Power, and Confidence Level
| Term | Meaning | Typical value |
|---|---|---|
| Confidence Level | How confident you want to be that your estimate contains the true population value | 95% |
| ╬▒ (alpha) | The significance level — your tolerance for a false positive (Type I error) = 1 − confidence level | 0.05 |
| Power (1−╬▓) | The probability of correctly detecting a real effect (avoiding a Type II error / false negative) | 80% |
| Effect size (d) | The size of the difference you consider meaningful, in standardized units | 0.5 (medium) |
ЁЯФв Calculator
ЁЯУР Calculation Steps
How to Use
- Choose the scenario that matches your research question from the dropdown.
- Fill in the required values — hover the hints under each field if you're unsure what to enter.
- Click Calculate. Tick "Show calculation steps" if you want to see exactly how the formula was applied — useful for including the working in a methodology chapter.
- Round the result up to the next whole number — you can always sample a few extra respondents to guard against non-response, but you can't safely sample fewer than the calculated minimum.
Practical Guidelines
- No pilot data for ╧Г or p? For proportions, use p = 0.5 (most conservative). For means, a rough estimate from similar published studies is standard practice — state this assumption in your methodology.
- Budget for non-response. If you expect a 20% non-response rate, divide your calculated n by 0.8 to get the number of people to actually approach.
- Smaller margin of error or higher confidence → larger required sample. These formulas show a well-known trade-off: precision costs sample size.
- These are planning estimates, not guarantees. They tell you the minimum size needed under your stated assumptions — if your pilot estimate of ╧Г or p turns out to be off, actual precision/power will differ.