The Central Limit Theorem
“The central limit theorem: add up enough of anything and it becomes a bell.”
The formula
(x̄ₙ − μ) / (σ/√n) → N(0, 1) (n → ∞)How to read it: whatever the original distribution, the averages of large samples follow a distribution that approaches the normal (bell) shape
- x̄ₙ
- — the mean of a sample of size n
- μ
- — the mean of the original distribution
- σ/√n
- — the standard error of the sample mean—the size of its wobble
- N(0,1)
- — the standard normal distribution (bell with mean 0, standard deviation 1)
The hook
One die is flat, but the sum of ten dice is bell-shaped. Here lies the reason the world is full of bell curves.
In plain words
The central limit theorem says that whatever the shape of the original distribution—flat or skewed—if you gather many independent samples and take their average (or sum), the distribution of those averages approaches the normal (bell) shape.
The intuition
Add up many small, independent random effects and the extremes (everything large or everything small) are rare while the middle (pushes and pulls cancel) is common. So the result is a bell, humped in the middle. It doesn't matter what shape each cause has—add enough of them and a bell appears. That's why heights, measurement errors, and dice sums are all bell-shaped.
How it's built
Two pieces—the center μ and the spread σ/√n. The sample means cluster around the true mean μ (law of large numbers), and the width of that cluster is the standard error σ/√n. As n grows the width narrows and the shape converges to a smooth normal curve. The theorem is about the shape of the distribution, not just where its center sits.
Example
One die gives a flat distribution—1 through 6 equally likely (mean 3.5). But roll ten dice and sum, and that sum piles up in a bell around 35. A flat input yields a bell-shaped output.
Common mistake
The theorem is about the distribution of the sample mean (or sum), not the raw data. The original data can be as skewed as you like—what becomes bell-shaped is the distribution of averages of many values, not the data itself.
Where it's used
The fundamental reason the normal distribution appears everywhere in statistics, margins of error in polls, confidence intervals, control charts in quality management, A/B testing, measurement-error analysis—nearly all inference from samples leans on this theorem.
Where it came from
De Moivre first spotted it for coin flips in 1733, Laplace generalized it, and Lyapunov and Lindeberg made it rigorous in the early 1900s. The 'central' means 'of central importance'.
Prerequisites
Quick check
By the central limit theorem, the distribution of the sample mean of many independent variables approaches which distribution?
- uniform (flat)
- normal (bell)✓
- skewed
- binomial
Practice
With standard deviation σ=10 and sample size n=100, find the standard error σ/√n of the sample mean.
Answer: 1
- √n = √100 = 10
- σ/√n = 10/10 = 1
Takeaway: The larger the sample, the smaller the wobble σ/√n of the mean.
With σ=20 and n=4, find the standard error σ/√n.
Answer: 10
- √n = √4 = 2
- σ/√n = 20/2 = 10
Takeaway: The standard error is inversely proportional to √n.
If the sample size n is increased fourfold, by what factor does the standard error of the mean shrink?
Answer: 2
- the standard error is σ/√n
- 4× n means √n doubles → the error halves (factor 2)
Takeaway: To double precision, quadruple the sample.
Explain with the central limit theorem why so many natural quantities (heights, measurement errors) are bell-shaped.
Answer: undefined
- such values are often the sum of many small, independent factors
- many small effects added together approach a normal distribution by the CLT
- so whatever the causes, the result comes out bell-shaped
Takeaway: Many small causes summed → a bell. That's the CLT's message.