4.15 Central limit theorem

4.15 Central limit theorem

IB Maths: Applications and Interpretation HL: Revision notes

Studying this course? Keep your progress in one place.

Key facts

  • Any linear combination of independent normal variables is normal; variances add, even for X−YX-Y.
  • The total of nn copies of X∼N(μ,σ2)X\sim N(\mu,\sigma^2) is N(nμ, nσ2)N(n\mu,\,n\sigma^2).
  • For a normal population, Xˉ∼N(μ,σ2n)\bar X\sim N\left(\mu,\frac{\sigma^2}{n}\right) for any nn; the standard error is σn\frac{\sigma}{\sqrt n}.
  • The central limit theorem: for large nn (n>30n>30 in IB), Xˉ\bar X is approximately normal whatever the population.
  • Decide first whether the question is about one item, a total or a mean.

Combining normal variables

A linear combination of independent normal variables is itself normal, so you only need its mean and variance.

If X1,…,XnX_1,\dots,X_n are independent normal variables, any linear combination a1X1+⋯+anXna_1X_1+\dots+a_nX_n is also normal.

The total of nn independent copies of X∼N(μ,σ2)X\sim N(\mu,\sigma^2) is N(nμ, nσ2)N(n\mu,\ n\sigma^2).

Variances add even when you subtract the variables.

−6−4−22460.10.20.30.40.5xyX ~ N(0, 1)X₁ + X₂ ~ N(0, 2)
For independent X₁, X₂ ~ N(0, 1), the sum is normal with the same mean and double the variance: N(0, 2), so it is wider and lower.
  • MeanE(aX±bY)=aE(X)±bE(Y)\mathrm{E}(aX\pm bY)=a\mathrm{E}(X)\pm b\mathrm{E}(Y)
  • Variance (independent)Var(aX±bY)=a2Var(X)+b2Var(Y)\mathrm{Var}(aX\pm bY)=a^2\mathrm{Var}(X)+b^2\mathrm{Var}(Y)

Worked example

Six adults each have mass N(78,112)N(78,11^2) kg, independently. Find the distribution of their total mass.

XX and YY are independent with Var(X)=4\mathrm{Var}(X)=4 and Var(Y)=9\mathrm{Var}(Y)=9. What is Var(X−Y)\mathrm{Var}(X-Y)?

Distribution of the sample mean

The sample mean keeps the mean μ\mu, but its variance shrinks by a factor of nn.

The sample mean is Xˉ=X1+⋯+Xnn\bar X=\frac{X_1+\dots+X_n}{n}. If X∼N(μ,σ2)X\sim N(\mu,\sigma^2) and the sample is random, then Xˉ∼N(μ, σ2n)\bar X\sim N\left(\mu,\ \frac{\sigma^2}{n}\right).

The standard deviation of Xˉ\bar X is σn\frac{\sigma}{\sqrt n}, the standard error. Larger samples give means that cluster more tightly around μ\mu.

This is exact when the population is normal, for any nn.

−3−2−11230.511.522.5xySample mean, size nSingle value X
Distribution of the sample mean for a standard normal population. Drag n: the curve gets narrower and taller.

Worked example

A machine fills bags with σ=0.04\sigma=0.04 kg. Find the standard deviation of the mean of n=9n=9 bags.

X∼N(50, 122)X\sim N(50,\ 12^2) and n=36n=36. What is the standard deviation of Xˉ\bar X?

The central limit theorem

For large nn, the sample mean is approximately normal whatever the shape of the population.

The central limit theorem says that, for any population with finite mean μ\mu and variance σ2\sigma^2, the distribution of Xˉ\bar X approaches N(μ,σ2n)N\left(\mu,\frac{\sigma^2}{n}\right) as nn becomes large.

A nearly symmetric population needs a small nn; a very skewed one needs a larger nn.

In examinations, n>30n>30 is considered sufficient. The result is an approximation unless the population is normal, so the sample mean is only approximately normal.

  1. 1

    Take a sample

    of size n from a skewed population

  2. 2

    Record its mean

    one value of the sample mean

  3. 3

    Repeat many times

    for thousands of samples

  4. 4

    Plot a histogram

    of all the sample means

  5. 5

    Increase n

    the histogram becomes bell-shaped

What a CLT simulation does

A population is strongly skewed. For which sample size can you use a normal model for the sample mean in IB?

Worked examples with the GDC

State the distribution of Xˉ\bar X with both parameters, say why it is valid, then use the GDC.

Always say why the normal model is valid: either the population is normal, or n>30n>30 and the CLT applies.

State the distribution of the sample mean, with both parameters, before using the GDC.

For a required sample size, use the inverse normal on the GDC, then round nn up.

44.555.566.577.588.590.10.20.30.40.50.60.7xy7Sample mean, N(6, 0.405)
Waiting time example: P(Tˉ>7)\mathrm{P}(\bar T>7) is the shaded upper tail.

Worked example

A call centre's waiting time TT has mean 66 and standard deviation 4.54.5 (skewed). For a sample of n=50n=50 calls, find P(Tˉ>7)\mathrm{P}(\bar T>7).

Worked example: sample size

σ=15\sigma=15. Find the smallest nn so that P(Xˉ<μ−3)<0.01\mathrm{P}(\bar X<\mu-3)<0.01.

Why is the CLT used in the call centre example?

Choosing the right approach

Decide whether the question is about one item, a total or a mean before choosing the parameters.

  • Population normal: Xˉ\bar X is exactly normal for any nn.
  • Population not normal and n>30n>30: Xˉ\bar X is approximately normal (CLT).
  • Population not normal and n≤30n\le30: no normal model is justified in IB.
  • A single value uses σ\sigma; a sample mean uses σn\frac{\sigma}{\sqrt n}.
−3−2−11230.20.40.60.81xyone item: N(0, 1)mean of 4: N(0, 0.25)
For X ~ N(0, 1), the mean of 4 values has standard deviation 1/√4 = 0.5, so its distribution is narrower and taller than a single value.

One item

Distribution:
N(μ, σ2)N(\mu,\ \sigma^2)
Standard deviation:
σ\sigma

Total of n

Distribution:
N(nμ, nσ2)N(n\mu,\ n\sigma^2)
Standard deviation:
σn\sigma\sqrt n

Mean of n

Distribution:
N(μ, σ2n)N\left(\mu,\ \frac{\sigma^2}{n}\right)
Standard deviation:
σn\frac{\sigma}{\sqrt n}

Masses of eggs are N(60, 42)N(60,\ 4^2). You want the distribution of the mean of 44 eggs. Which is correct?

Try an exam question

The mass of an apple is N(150, 122)N(150,\ 12^2) g. Four apples are chosen at random. (a) Find the distribution of their total mass. (b) Find the probability that the total mass is more than 620620 g.

[4 marks]

That's the notes covered.

Carry on to the next subtopic.

Try a question

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

1 mark, marked against the mark scheme

Revision notes