Watch the mean take shape.
One replication = draw n independent observations and record their mean. Repeat to reveal the sampling distribution.
Your experiment
Changing shape or n starts a new experiment. Repeated runs add results (up to 100,000 means).Population distribution
Distribution of sample means
More observations. Less uncertainty.
The spread of sample means is the standard error (SE). Quadruple n to cut SE in half.
Precision controls
The square-root relationship
| n | SE | SE as % of σ |
|---|
A person is not an average.
Boeing seat-width example: compare an individual hip width with the mean hip width of 126 passengers.
Normal probability calculator
Choose the right denominator
These calculations assume a normal population and independent random observations. If the population is nonnormal, only the mean may be approximately normal when the sample is sufficiently large; the CLT does not make individual values normal.
How surprising is 98.2°F?
Assume the population mean is 98.6°F. Find the probability of a sample mean at or below the observed value.
Body-temperature example
A probability under an assumption
This left-tail probability measures how unusual this result or a lower mean would be if the assumed population model were correct. It is not the probability that the assumed mean is true.
Use a random, representative sample with independent measurements and known population σ. Normality of the sample mean is exact for a normal population and approximate otherwise. The example uses a large-sample approximation; severe skewness, dependence or heavy tails can undermine it. A one-sided direction should be chosen before examining the data.
When the population is finite.
Sampling without replacement removes uncertainty faster as the sample covers more of the population.
Population & sample
For a simple random sample without replacement, with σ² = Σ(xᵢ − μ)² / N. If population variance is defined using N − 1 instead, use SE = (S / √n) √(1 − n/N).
Boundary checks
At n = 1 the correction is 1. At n = N a census has SE = 0. With replacement, draws are independent and SE = σ / √n, even for a finite population.
The notes recommend applying the correction when n exceeds 5% of N. The correction formula is valid at any permitted fraction without replacement; 5% is a convention for deciding when ignoring it has little effect. It corrects the variance, but does not guarantee a normal sampling distribution.
The theorem, assumptions & how to use this explorer
For independent, identically distributed observations with finite mean μ and finite variance σ², the standardized sample mean approaches the standard normal distribution as n increases. The mean of X̄ is μ and its standard deviation is σ/√n under these assumptions, whether or not X̄ is already approximately normal.
A normal population gives an exactly normal sample mean for every n. For other populations, the speed of convergence depends on skewness and tail behavior. “n > 30” is a classroom guideline, not a universal cutoff. Some symmetric populations work well at smaller n; strongly skewed or heavy-tailed populations may need much larger n. Infinite-variance populations are outside this classical CLT.
Simulation uses independent random draws (equivalent to sampling with replacement). Histograms show density, so total bar area is 1; the orange curve is N(μ, σ²/n), a comparison rather than an assertion of exact normality. All simulations are generated on your device and will vary between runs.
Source: Statistics I, Chapter 6 Normal Probability Distributions, 6-4 The Central Limit Theorem, 14-page course notes, “6-4 The Central Limit Theorem”. Tabs 03–04 reproduce the examples on pages 9 and 11–12; tab 05 uses page 13. The weight-gain preset follows page 14. Historical aircraft details and temperature values are used as course exercises, not current design or health guidance.
Open this file in a modern browser. No installation, network, CDN, account or external library is needed. Use Tab to navigate controls; use arrow keys in the tab strip. Browser Print includes all five topics.