Choose a proportion or mean precision model and supply a positive target margin. An optional population must be a whole positive number; blank means infinite population, while zero, fractions and malformed tokens are rejected.
Read the result
Keep the continuous estimate n0 through finite-population correction, then take one ceiling: ceil(n0·N/(n0+N−1)). The correction assumes appropriate sampling without replacement. The n and n−1 achieved margins demonstrate the integer boundary within the stated approximation.
Precision and detecting an effect are different questions. Native proportion power uses explicit p0 and p1 in a two-sided one-sample normal approximation. Mean power uses a specified difference and SD. The standardized reference curves are independent; quick-fill clinical or A/B labels do not specify a full trial.
Conventions and limits
Normal approximations need model checks; n≥30 alone is no guarantee. Mean sizing below 30 uses an approximate t correction. SD from range/4 is rough planning, not an identified distribution. All fixed-n tools and assumptions restore together; invalid margins clear dependent scenarios and power results.
Compare assumptions before choosing n
The population-comparison workspace holds p and the target margin fixed while varying N. Its declared fixed z=1.96 convention corresponds approximately to 95% confidence; the native calculator derives its critical value from the selected confidence. Read the achieved margins at n and n−1 to see the integer boundary. State whether the calculation targets precision or detection power before comparing plans.