How this instrument works
The Central Limit Theorem says that the distribution of sample means — if you were to repeatedly draw samples of size n from a population and compute each sample's average — is itself approximately normal, centered on the population mean μ, with a standard deviation smaller than the population's own σ. That smaller spread is the standard error, SE = σ/√n: the larger the sample size, the more tightly sample means cluster around the true population mean, since averaging tends to smooth out individual extremes.
This is a distinct question from asking about a single raw observation. This site's normal-distribution and normal-approximation calculators work with individual values drawn straight from a distribution with standard deviation σ; this calculator instead works with the average of n such values, which has its own, smaller standard deviation of σ/√n. Using the raw population σ instead of the standard error when evaluating a sample mean is a common and consequential mistake — it overstates how much a sample mean should vary.
Once you have the standard error, the same z-score and normal-CDF machinery applies: standardize the sample mean using z = (x̄−μ)/SE, then look up the cumulative probability. Because SE shrinks as n grows, the same gap between x̄ and μ becomes a larger, more surprising z-score, and a more extreme probability, as the sample size increases — exactly why larger samples make it easier to detect a real difference from the population mean.
- Enter the population mean into μ (mu).
- Enter the population standard deviation into σ (sigma) — the raw population spread, not a standard error.
- Enter the sample size into n — the number of observations averaged to form the sample mean.
- Enter the sample-mean value being evaluated into x̄ (xbar).
- Read SE, the standard error σ/√n, z, x̄'s distance from μ in standard-error units, and the resulting probability.
Worked example — a sample mean from n = 25 draws
A population has mean μ = 100 and standard deviation σ = 15. You draw a sample of n = 25 observations and want the probability the sample mean is 105 or less. Enter 100 into μ, 15 into σ, 25 into n, and 105 into x̄.
SE = 15/sqrt(25) = 15/5 = 3.0 exactly — the sample mean's own standard deviation is one-fifth the size of a single raw observation's. z = (105−100)/3 = 1.666667. P(x̄ ≤ 105) ≈ Phi(1.666667) ≈ 0.9522, meaning about 95.2% of samples of size 25 from this population would produce a mean of 105 or less — a notably tighter, more confident statement than could be made about any single raw observation from the same population.
Questions
How is this different from a normal-distribution calculation on a single value?
A single raw value's spread is the population standard deviation, σ, directly. A sample mean's spread — since averaging multiple observations smooths out individual extremes — is smaller: the standard error, SE = σ/√n. Plugging the raw σ into a z-score calculation meant for a sample mean, instead of the smaller SE, is a common mistake that understates how unusual a given sample mean actually is.
Why does a larger sample size make the same sample mean more unusual?
Because the standard error shrinks as n grows, divided by the square root of n, the same numeric gap between the sample mean and the population mean corresponds to a larger z-score, and therefore a more extreme probability, with a bigger sample. A large sample's average is expected to land very close to the true population mean purely by the law of large numbers, so a noticeable gap becomes harder to explain away as ordinary sampling variation once the sample is large.
Does this rely on the population itself being normally distributed?
Not strictly — that's the point of the Central Limit Theorem. Even if the underlying population's distribution is skewed or otherwise non-normal, the distribution of sample means becomes approximately normal as n grows, typically by around n = 30 for most reasonably well-behaved populations. For very small samples from a strongly non-normal population, the normal approximation used here may be less accurate.
What happens if n = 1?
With n = 1, the standard error collapses to SE = σ/√1 = σ, and the sample mean is just the single raw observation itself, so this calculator's result becomes mathematically identical to evaluating that one value directly against the population's normal distribution. It's a useful sanity-check case: sampling-distribution calculations should always reduce to the ordinary single-value case when the sample size is 1.