The Central Limit Theorem

The Central Limit Theorem

Spec 5.1 · Edexcel A-Level Further Maths: Revision notes

Studying this course? Keep your progress in one place.

Key facts

  • For a large sample size nn, the sample mean is approximately normal: Xˉ≈N(μ,σ2n)\bar{X}\approx\mathrm{N}\left(\mu,\frac{\sigma^2}{n}\right).
  • It holds whatever the shape of the population; the sample mean, not the population, is the one that is normal.
  • Find μ\mu and σ2\sigma^2 for one observation first, then divide σ2\sigma^2 by nn.
  • Standardise with z=xˉ−μσ/nz=\frac{\bar{x}-\mu}{\sigma/\sqrt n}; σ/n\sigma/\sqrt n is the standard error.
  • For a smallest sample size, solve with the inverse normal and round nn up.

What the CLT says

For a large sample, the sample mean is approximately normal, whatever the population's shape.

The Central Limit Theorem (CLT) says that if a population has mean μ\mu and variance σ2\sigma^2, then for a large sample size nn the sample mean Xˉ\bar{X} is approximately normally distributed.

This holds whatever the shape of the population distribution, so the population does not have to be normal. You can use it for Poisson, binomial, geometric and negative binomial populations. You do not need to prove it.

The larger nn is, the better the approximation; about 3030 or more is usually accepted as large. Saying the CLT makes the population normal is a common slip.

−3−2−11230.511.52xdensitypopulation N(0, 1)sample mean N(0, 1/n)
Illustrative: a population N(0, 1) and the distribution of the sample mean, which narrows as n grows. Move the slider.
  • Large nnXˉ≈N(μ,σ2n)\bar{X}\approx\mathrm{N}\left(\mu,\dfrac{\sigma^2}{n}\right)

What does the Central Limit Theorem say is approximately normal?

Population mean and variance

Use the mean and variance of a single observation from the population, from the standard results.

To use the CLT you need μ\mu and the population variance σ2\sigma^2 for a single observation. Use the standard results.

Be careful with the symbol nn. For a binomial population, nn in B(n,p)\mathrm{B}(n,p) is the number of trials in one observation, whereas the CLT sample size is the number of observations.

Write down the population μ\mu and σ2\sigma^2 before anything else, then divide σ2\sigma^2 by the sample size to get the variance of Xˉ\bar{X}.

510152025300.20.40.60.81nsd of X̄n = 4: 0.5n = 25: 0.2σ/√n, σ = 1
The standard deviation of the sample mean is σ/√n (shown with σ = 1), so it falls as the sample grows.

Poisson Po(λ)

Mean μ\mu:
λ\lambda
Variance σ2\sigma^2:
λ\lambda

Binomial B(n, p)

Mean μ\mu:
npnp
Variance σ2\sigma^2:
np(1−p)np(1-p)

Geometric Geo(p)

Mean μ\mu:
1p\frac1p
Variance σ2\sigma^2:
1−pp2\frac{1-p}{p^2}

Negative binomial NB(r, p)

Mean μ\mu:
rp\frac rp
Variance σ2\sigma^2:
r(1−p)p2\frac{r(1-p)}{p^2}

Worked example

A box holds 1212 pens, each faulty with probability 0.250.25. A sample of 4040 boxes is taken. State the approximate distribution of the sample mean number of faulty pens.

For X∼Geo(0.5)X\sim\mathrm{Geo}(0.5), what is σ2\sigma^2?

Probabilities for a sample mean

Find μ\mu and σ2\sigma^2, state the CLT distribution, standardise with the standard error and use the normal tables.

Follow these steps: find μ\mu and σ2\sigma^2 of one observation; state Xˉ≈N(μ,σ2/n)\bar{X}\approx\mathrm{N}(\mu,\sigma^2/n) and say the CLT is used because nn is large; standardise; read the probability from the normal distribution.

The standard error is σn\frac{\sigma}{\sqrt n}, the standard deviation of Xˉ\bar{X}. A sample mean varies far less than a single observation. For a mean within aa of μ\mu: P(μ−a<Xˉ<μ+a)=2Φ(aσ/n)−1\mathrm{P}(\mu-a<\bar{X}<\mu+a)=2\Phi\left(\frac{a}{\sigma/\sqrt n}\right)-1.

  1. 1

    Population

    Find μ\mu and σ2\sigma^2 of one observation

  2. 2

    State

    Xˉ≈N(μ,σ2/n)\bar{X}\approx\mathrm{N}(\mu,\sigma^2/n), because nn is large

  3. 3

    Standardise

    z=xˉ−μσ/nz=\frac{\bar{x}-\mu}{\sigma/\sqrt n}

  4. 4

    Read

    Use the normal distribution on your calculator

Using the CLT.

Worked example

X∼Po(4)X\sim\mathrm{Po}(4) and a sample of n=50n=50 is taken. Find P(Xˉ>4.5)\mathrm{P}(\bar{X}>4.5).

A population has σ=10\sigma=10 and n=25n=25. What is the standard deviation of Xˉ\bar{X}?

Sample size and when CLT applies

Standardise with nn unknown, use the inverse normal and round nn up.

For the smallest nn so that a probability condition holds, standardise with nn unknown and use the inverse normal. Always round nn up to the next whole number.

To justify the CLT, state that the sample is large, so the sample mean is approximately normal even though the population is not. The approximation is poor for small nn when the population is very skewed, such as a geometric distribution with small pp. The observations must also be independent.

In an "explain" question, name the Central Limit Theorem and say why it applies.

1020304050600.511.52nzn = 43.3, round up to 44z = k√n/σz = 1.645
Illustrative: σ = 2, k = 0.5. The z-value k√n/σ must exceed 1.645, which first happens at n = 43.3, so n = 44.
  • kσ/n>z\dfrac{k}{\sigma/\sqrt n}>zn>(zσk)2n>\left(\dfrac{z\sigma}{k}\right)^2

Worked example

X∼NB(3,0.25)X\sim\mathrm{NB}(3,0.25). Find the smallest nn such that P(Xˉ>13)<0.01\mathrm{P}(\bar{X}>13)<0.01.

A solution gives n>194.8n>194.8. What sample size do you state?

Try an exam question

A random variable X∼B(10,0.3)X\sim\mathrm{B}(10,0.3). A random sample of 4040 observations of XX is taken, with sample mean Xˉ\bar{X}. Use the Central Limit Theorem to find P(Xˉ>3.5)\mathrm{P}(\bar{X}>3.5).

[5 marks]

That's the notes covered.

Carry on to the next subtopic.

Try a question

Which of the following is the approximate distribution of Xˉ\bar{X}?

1 mark, marked against the mark scheme

Revision notes