Direct answer
For common independent-sample estimators, margin of error equals a critical value times standard error; reducing it usually requires more observations, lower variability, or a different design, with required sample sizes rounded upward.
Visual explanation
Precision improves with diminishing returns
What this calculation tells you
Margin of error describes sampling precision under one interval procedure. Sample-size planning reverses that relationship using an anticipated variability or proportion and a chosen target precision.
Define the estimand, acceptable precision, confidence level, design, expected variability, nonresponse allowance, analysis model, and feasibility before calculating n.
Where it is used
Surveys
Plan a simple precision target while documenting design limitations.
Experiments
Connect effect, variability, power, and allocation before recruitment.
Quality
Plan measurement effort around a stated precision target.
Education
Demonstrate square-root returns to sample size.
Common situations
- Planning a mean or proportion estimate.
- Comparing 90%, 95%, and 99% confidence.
- Allowing for design effect or attrition.
- Explaining why doubling n does not halve error.
Start with the statistical question
Define the estimand, acceptable precision, confidence level, design, expected variability, nonresponse allowance, analysis model, and feasibility before calculating n.
Sliders for n, variability, and confidence change an interval width. A square-root curve shows diminishing precision gains as sample size grows.
Worked example
Under a simple mean model with fixed spread, SE is proportional to 1/√n. Moving from n=100 to n=400 halves SE; moving from 100 to 200 reduces it only by √2.
Assumptions that carry the result
Closed-form plans rely on anticipated parameters and simple designs. Clustering, weighting, attrition, unequal allocation, repeated measures, finite populations, and planned analyses change the requirement.
Interpret the result without overreaching
A large n cannot correct selection bias or poor measurement. A sample-size result is a planning input, not a guarantee of power, precision, recruitment, or scientific value.
- Choosing n from population size alone.
- Rounding required sample size down.
- Ignoring nonresponse, clustering, or the planned analysis.
Practical questions
Frequently asked questions
Does a bigger population always require a bigger sample?
Not under ordinary large-population simple-random precision formulas; finite-population correction matters when the sampling fraction is substantial.
Why round sample size upward?
Rounding down would fail to meet the calculated requirement under the planning approximation.
Can I plan without a variance estimate?
You need a justified conservative value, pilot estimate, prior evidence, or sensitivity analysis rather than pretending variability is known.
Further reading
Authoritative sources
Use these primary and professional resources to check definitions, conventions, or requirements that may extend beyond this guide.