Take a lopsided population — incomes, wait times, anything skewed — average a random sample of it, and the average behaves beautifully. Repeat with bigger samples and the averages cluster into a bell curve centered exactly on the population mean, with a spread you can compute before collecting a single data point. That is the central limit theorem, the result half of statistics stands on. This central limit theorem calculator turns a population mean, population standard deviation, and sample size into the sampling distribution — and the probability that a sample mean lands below or above any value you choose.
What the Central Limit Theorem Says
Draw random samples of size n from any population with mean μ and standard deviation σ, and compute each sample’s mean. Those sample means form their own distribution — the sampling distribution of the mean — and the theorem guarantees three things: its mean is exactly μ, its standard deviation is σ ÷ √n (the standard error), and its shape approaches normal as n grows, whatever the population’s shape. The theorem is why normal-based methods work on stubbornly non-normal data: you are rarely modeling raw values, you are modeling averages, and averages go normal.
The Standard Error: Sigma Over Root N
The formula σx̄ = σ ÷ √n carries the practical punch. Averaging cancels noise: a sample of 36 compresses the spread sixfold, because √36 = 6. To halve the standard error again you need n = 144 — the square-root law that governs the cost of precision everywhere from opinion polls to clinical trials. Note the standard error describes the scatter of sample means, not of individual values; confusing the two is the most common misreading of the theorem.
Worked Example: Sample Means of an Exam
Exam scores in a large population have μ = 100 and σ = 15. Take samples of n = 36 students. The standard error is 15 ÷ 6 = 2.5. What is the probability a class of 36 averages below 105? Standardize: z = (105 − 100) ÷ 2.5 = 2.00, and Φ(2.00) = 0.97725 — a 97.72 percent chance the sample mean lands under 105, and only 2.28 percent that a class averages above it. Individual students score above 105 about 37 percent of the time; whole classes almost never do. That contrast is the theorem in action.
How Large Must N Be?
The textbook rule is n ≥ 30 for the normal approximation to serve, but the honest answer is “it depends on the population’s shape.” For a symmetric, tame population, n = 5 can look convincingly normal. For a heavily skewed one — lottery payouts, insurance claims — even n = 50 may still lean. The theorem promises convergence, not a schedule. When the population itself is normal, no threshold is needed at all: sample means are exactly normal at every n, even n = 1. When σ is unknown and estimated from the sample, small samples call for the t-distribution’s fatter tails instead — the bridge between this theorem and practical inference.
Why Polls, Trials, and Factories Run on This Theorem
A poll’s margin of error is a standard error in a suit. A clinical trial’s power calculation starts from the sampling distribution of a mean difference. A factory’s control chart plots sample means against limits set at μ ± 3 standard errors — limits that only make sense because those means are approximately normal even when individual parts are not. Remove the central limit theorem and none of these tools has a foundation; with it, the behavior of averages is predictable before data arrives.
Common Mistakes
Claiming the theorem makes the data normal — it does not; it normalizes the distribution of the mean. Using σ where σ ÷ √n belongs (or the reverse) flips answers by a factor of √n. Applying the approximation at n = 4 to a wildly skewed population and trusting the third decimal. And forgetting the fine print: samples must be random and independent — a biased sample’s means converge beautifully to the wrong center.
Related Tools
Continue in the Statistics Calculators hub with the P-Hat Calculator — the same theorem applied to proportions — and the Z-Score Calculator.
Frequently Asked Questions
What is the central limit theorem in simple terms?
Averages of random samples follow a normal distribution centered on the population mean, with standard deviation sigma divided by the square root of the sample size — no matter what shape the original population has, as the sample grows.
How do you calculate the standard error of the mean?
Divide the population standard deviation by the square root of the sample size: sigma ÷ √n. For sigma = 15 and n = 36, the standard error is 15 ÷ 6 = 2.5.
What sample size is large enough for the CLT?
The common rule is n of at least 30, but shape matters: nearly normal populations need far fewer, severely skewed populations need more. If the population is itself normal, sample means are normal at any n.
What is the probability a sample mean exceeds a value?
Standardize with the standard error: z = (x-bar − mu) ÷ (sigma ÷ √n), then read the normal curve. This calculator’s worked example: a mean of 105 with mu 100, sigma 15, n 36 gives z = 2.00 and only a 2.28 percent chance of exceeding it.
Does the CLT apply to proportions too?
Yes. A sample proportion is a mean of zeros and ones, so p-hat is approximately normal with standard error √( p(1−p) ÷ n ) — the machinery behind the P-Hat Calculator.
Why does averaging make things normal?
Extremes cancel: a sample mean is extreme only if most of its values lean the same way, which gets rarer as n grows. The mathematics (characteristic functions, in the formal proof) confirms the intuition — sums of independent random quantities converge to the normal shape.
What if sigma is unknown?
Estimate it with the sample standard deviation s. For large samples the normal calculation barely changes; for small ones, switch to the t-distribution, whose heavier tails price in the extra uncertainty.