Confidence Interval Calculator

Mean (known σ) / Proportion (Wald); 90%, 95%, 99%, 99.9%

Lower limit
—
Upper limit
—
Margin of error
—
Point estimate
—
Standard error
—
Critical z
—

Two-sided normal intervals. Mean mode assumes independent observations, a normally distributed sample mean and known population σ; it is not a Student t interval. Proportion mode uses the Wald approximation, which can have poor coverage for small samples or proportions near 0/1; endpoints are not clipped to [0,1]. Confidence is long-run coverage, not a probability assigned to this fixed interval. Presets use precise normal critical values; custom z can reproduce rounded source values. Inputs: |mean|,σ≤10¹², integer n=1–10¹², p=0–1, custom z>0 and ≤37; reduced decimal denominator ≤10²⁴. Use a decimal point or comma; inactive fields are ignored. Results are approximate to 12 significant digits.

How the calculation works

x̄ ± zσ/√n; p̂ ± z√(p̂(1−p̂)/n)

Two-sided normal intervals. Mean mode assumes independent observations, a normally distributed sample mean and known population σ; it is not a Student t interval. Proportion mode uses the Wald approximation, which can have poor coverage for small samples or proportions near 0/1; endpoints are not clipped to [0,1]. Confidence is long-run coverage, not a probability assigned to this fixed interval. Presets use precise normal critical values; custom z can reproduce rounded source values. Inputs: |mean|,σ≤10¹², integer n=1–10¹², p=0–1, custom z>0 and ≤37; reduced decimal denominator ≤10²⁴. Use a decimal point or comma; inactive fields are ignored. Results are approximate to 12 significant digits.

Frequently asked questions

How is the result calculated?

x̄ ± zσ/√n; p̂ ± z√(p̂(1−p̂)/n)

What assumptions apply?

Two-sided normal intervals. Mean mode assumes independent observations, a normally distributed sample mean and known population σ; it is not a Student t interval. Proportion mode uses the Wald approximation, which can have poor coverage for small samples or proportions near 0/1; endpoints are not clipped to [0,1]. Confidence is long-run coverage, not a probability assigned to this fixed interval. Presets use precise normal critical values; custom z can reproduce rounded source values. Inputs: |mean|,σ≤10¹², integer n=1–10¹², p=0–1, custom z>0 and ≤37; reduced decimal denominator ≤10²⁴. Use a decimal point or comma; inactive fields are ignored. Results are approximate to 12 significant digits.

How precise are the displayed results?

Two-sided normal intervals. Mean mode assumes independent observations, a normally distributed sample mean and known population σ; it is not a Student t interval. Proportion mode uses the Wald approximation, which can have poor coverage for small samples or proportions near 0/1; endpoints are not clipped to [0,1]. Confidence is long-run coverage, not a probability assigned to this fixed interval. Presets use precise normal critical values; custom z can reproduce rounded source values. Inputs: |mean|,σ≤10¹², integer n=1–10¹², p=0–1, custom z>0 and ≤37; reduced decimal denominator ≤10²⁴. Use a decimal point or comma; inactive fields are ignored. Results are approximate to 12 significant digits.

Also try:

Back to category →