Skip to main content

Sample Size Calculator

Wald sample size for a proportion. n = z^2 p (1-p) / e^2. Optional finite N. Educational, not a survey product and not a power analysis for a mean.

Rated 4.7 out of 5 based on 329 reviews

Results

How to compute a Wald sample size

  1. Enter z greater than 0. 1.96 is the usual two-sided 95% normal value. You type z. This page does not look up a table.
  2. Enter proportion p from 0 to 1. 0.5 maximizes p(1-p) and is the conservative default.
  3. Enter margin of error e greater than 0 as a proportion, so 5 percentage points is 0.05.
  4. Optional: type a finite population N greater than 1. Read n, n rounded up, and the finite rows when N is set. Educational only.

Wald sample size for a proportion

Educational n from z, p, and margin e

Formulan = z^2 p (1-p) / e^2
Defaultz = 1.96, p = 0.5, e = 0.05
Finite correctionn / (1 + (n-1)/N) when N > 1
Rounded nceil of the Wald n
z sourceYou type z. No inverse-normal table is loaded.
ScopeEducational Wald. Not a survey product.

What this sample size page computes

The Wald formula sizes a simple random sample for a binary proportion at a chosen z and margin of error. It is a classroom identity, not Cochran with finite fpc as a product default, and not a clustered or stratified design. Power for a mean, an odds ratio, or a two-sample test is out of scope.

How n is computed

n = z^2 p (1-p) / e^2. z and e must be greater than 0. p must sit from 0 to 1 inclusive. The page also ceils n. If you type a population N greater than 1, finite n is n / (1 + (n-1)/N), then that value is ceiled too. Convert a confidence level to z yourself. The z-score page standardizes one observation. It does not invert a normal tail into z for this form.

Wald, p = 0.5, and related z jobs

p = 0.5 is the usual conservative pick because p(1-p) peaks there. Wilson and Agresti-Coull intervals change how you report a proportion after you sample. They do not change this n formula. Design effect, nonresponse, and finite surveys with complex weights need a statistician. This page stays on the Wald identity.

Use cases

A methods-class check that 1.96, 0.5, and 0.05 produce n near 385. A planning sketch before you talk to a survey vendor. Not a replacement for IRB sample-size language or a vendor quote.

Limits

No inverse-normal lookup. No clusters. No two-sample power. Blank N skips finite correction. Educational only. CZNull does not receive the numbers.

Sample size questions

What is the formula?
n = z^2 p (1-p) / e^2. The default 1.96, 0.5, and 0.05 is the usual 95% Wald sketch.
Is this a survey product?
No. It is an educational Wald figure. Vendors add design effect and nonresponse.
Why is p default 0.5?
p(1-p) is largest at 0.5, so that choice gives the largest n for a given z and e.
How do I get z from 95%?
Type 1.96 if that is the quantile you want. This page does not convert a percent into z.
When does finite N apply?
When you enter a population greater than 1. The page then reports the corrected n and its ceil.
Does this size a t-test for a mean?
No. The formula is for a proportion. Mean power needs a different identity.