"How many respondents do I need?" is the question most often answered by copying a number from a senior's thesis. It is also the number external examiners most reliably ask you to justify, so a figure you cannot derive is a figure that will cost you in the viva.
This calculator uses Cochran's formula with the finite population correction — the method cited in most TU, PU and KU methodology chapters. It shows the arithmetic and the reasoning next to the result so you can defend the number rather than merely state it.
n₀ = z²·p·(1−p) / e². The z value comes from the confidence level (1.96 for 95%), p is the expected proportion and e is the margin of error as a decimal. At 95% confidence, ±5% and p = 0.5, the result is 385.
Where the population size N is known, the finite population correction applies: n = n₀ / (1 + (n₀−1)/N). For a population of 1,000 this reduces 385 to 278. That matters for the kind of bounded populations Nepali student research usually works with — one campus, one ward, one hospital, one cooperative — where the uncorrected figure is needlessly large.
The margin of error is the width of your answer: ±5% means a finding of 60% could reasonably lie between 55% and 65%. The confidence level is how often that interval would capture the true value across repeated samples — 19 times in 20 at 95%.
Tightening either costs sample size, and not linearly. Halving the margin from ±5% to ±2.5% quadruples the sample, because e is squared. This is why ±5% at 95% is close to universal in student work: it is where precision and feasible fieldwork meet.
The term p·(1−p) is the variance of a proportion and peaks at p = 0.5, so 50% produces the largest sample the formula will ask for. That makes it the cautious default when you have no prior estimate.
If a previous study or a national survey — a CBS report, a DHS round, an NRB survey — gives you a defensible figure, using it will reduce the required sample. Only do so when you can cite the source, since an optimistic guess quietly weakens your study. Entering 0% or 100% implies no variability at all, so this tool substitutes 50% and says that it has.
The calculated figure is the number of usable responses you need, not the number of people to contact. Needing 278 responses at an expected 70% response rate means approaching about 398.
Response rates vary sharply with method in Nepal: face-to-face administration in a ward often returns very high rates, while online forms circulated through social media frequently return far less. Base your figure on a pilot or on comparable local studies, and record it in your methodology. Finding the shortfall after fieldwork closes is expensive and sometimes unfixable.
It sizes a sample for estimating one proportion from a probability sample. If you intend to compare subgroups, each subgroup needs enough cases of its own: 278 responses spread across seven districts leaves about 40 each, too few for most tests. Size for the comparison you actually plan to make.
It also assumes probability sampling. Convenience and snowball samples cannot support a margin of error however many people you recruit, because the mathematics depends on each member of the population having a known chance of selection. Where your sampling is non-probability — as much student fieldwork necessarily is — report the sample size, drop the precision claim, and state the limitation honestly. For sizing an experiment, a power analysis is the correct tool.
278, at 95% confidence with a ±5% margin of error and 50% expected proportion. Changing any of those three inputs changes the answer.
It is the figure for 95% confidence, ±5% margin of error and maximum variability with a large or unknown population. If your population is bounded and known, the corrected figure is smaller.
50%, unless a prior study or national survey gives you a defensible estimate. 50% maximises variance and gives the largest, safest sample.
As a target, yes — but a margin of error is only meaningful for probability sampling. With a convenience sample, report the size and acknowledge the limitation.
Yes. Enter your expected response rate and the tool shows how many people to approach to end up with the sample you need.