Lesson 2 of 8 · 14 min

The mean and spread of sample means

Use E(Xˉ)=μE(\bar{X}) = \mu and SD(Xˉ)=σn\mathrm{SD}(\bar{X}) = \frac{\sigma}{\sqrt{n}}, see why larger samples give less variable means, and write Xˉ∼N(μ,σ2n)\bar{X} \sim N\left(\mu, \frac{\sigma^2}{n}\right) when the population is normal.

What you will be able to do

  • State and use E(Xˉ)=μE(\bar{X}) = \mu and SD(Xˉ)=σn\mathrm{SD}(\bar{X}) = \frac{\sigma}{\sqrt{n}}.
  • Predict how the spread of Xˉ\bar{X} changes when the sample size changes.
  • Write the distribution of Xˉ\bar{X} as N(μ,σ2n)N\left(\mu, \frac{\sigma^2}{n}\right) when XX is normal, with the variance in the second place.

A battery maker says its AA cells last a mean of 480 hours with standard deviation 36 hours, and lifetimes are normally distributed. One battery could easily last 440 or 520 hours. But the mean of 16 batteries? High and low lifetimes cancel out, so the mean should land much closer to 480. How much closer?

Interactive

How the spread of x̄ depends on n

Draw many samples with n=4n = 4, then n=16n = 16 and n=64n = 64. Read the standard deviation of the sample means each time. What happens when you multiply nn by 4?

The standard deviation of Xˉ\bar{X} for battery lifetimes with σ=36\sigma = 36 hours.
Sample size nn141664144
SD(Xˉ)=36n\mathrm{SD}(\bar{X}) = \frac{36}{\sqrt{n}}361894.53

Key idea

Mean and standard deviation of the sample mean

For random samples of size nn from a population with mean μ\mu and standard deviation σ\sigma:

E(Xˉ)=μSD(Xˉ)=σnVar(Xˉ)=σ2nE(\bar{X}) = \mu \qquad \mathrm{SD}(\bar{X}) = \frac{\sigma}{\sqrt{n}} \qquad \mathrm{Var}(\bar{X}) = \frac{\sigma^2}{n}

The sample mean is centred on the right value, and its spread shrinks as nn grows. σn\frac{\sigma}{\sqrt{n}} is called the standard error of the mean.

Why σ/√n and not σ/n?

Write Xˉ=1n(X1+X2+⋯+Xn)\bar{X} = \frac{1}{n}(X_1 + X_2 + \cdots + X_n), where the XiX_i are independent and each has mean μ\mu and variance σ2\sigma^2.

Mean. Expected values add, and E(aY)=aE(Y)E(aY) = aE(Y), so

E(Xˉ)=1n(μ+μ+⋯+μ)=1n(nμ)=μE(\bar{X}) = \frac{1}{n}(\mu + \mu + \cdots + \mu) = \frac{1}{n}(n\mu) = \mu

Variance. For independent random variables, variances add (the simulation above confirms it). And Var(aY)=a2 Var(Y)\mathrm{Var}(aY) = a^2\,\mathrm{Var}(Y), so

Var(Xˉ)=1n2(σ2+σ2+⋯+σ2)=1n2(nσ2)=σ2n\mathrm{Var}(\bar{X}) = \frac{1}{n^2}(\sigma^2 + \sigma^2 + \cdots + \sigma^2) = \frac{1}{n^2}(n\sigma^2) = \frac{\sigma^2}{n}

Taking the square root gives SD(Xˉ)=σn\mathrm{SD}(\bar{X}) = \frac{\sigma}{\sqrt{n}}. The variance is divided by nn, so the standard deviation is divided by n\sqrt{n}.

A population has mean 6060 and standard deviation 1212. Random samples of size 3636 are taken. Find the standard deviation of the sample mean Xˉ\bar{X}.

Key idea

When the population is normal

If X∼N(μ,σ2)X \sim N(\mu, \sigma^2), then the mean of a random sample of any size nn is exactly normal too:

Xˉ∼N(μ,σ2n)\bar{X} \sim N\left(\mu, \frac{\sigma^2}{n}\right)

As always in this notation, the second number is the variance. For the batteries with n=16n = 16: Xˉ∼N(480,36216)=N(480,81)\bar{X} \sim N\left(480, \frac{36^2}{16}\right) = N(480, 81), which has standard deviation 99.

One battery, X∼N(480,362)X \sim N(480, 36^2), against the mean of 16 batteries, Xˉ∼N(480,92)\bar{X} \sim N(480, 9^2). Same centre, a quarter of the spread.

Worked example

Mean, standard deviation and distribution of Xˉ\bar{X}

  • Simple familiar
  • Technology free

The lifetimes of a brand of AA battery are normally distributed with mean 480480 hours and standard deviation 3636 hours. Random samples of 1616 batteries are tested and the sample mean lifetime Xˉ\bar{X} is recorded.

(a) Find the mean and standard deviation of Xˉ\bar{X}.

(b) State the distribution of Xˉ\bar{X}.

(c) How many batteries should each sample contain for the standard deviation of Xˉ\bar{X} to be 4.54.5 hours?

4 steps

Try it yourself first, then check the working one step at a time.

Method

  1. E(Xˉ)=μE(\bar{X}) = \mu and SD(Xˉ)=σn\mathrm{SD}(\bar{X}) = \frac{\sigma}{\sqrt{n}}.

  2. Write Xˉ∼N(μ,σ2n)\bar{X} \sim N\left(\mu, \frac{\sigma^2}{n}\right) and say why it is normal.

  3. To reach a target standard deviation, solve σn=target\frac{\sigma}{\sqrt{n}} = \text{target} for nn.

Watch out for

  • Writing N(480,9)N(480, 9): the second parameter is the variance, 8181.

  • Dividing by nn instead of n\sqrt{n}, giving 2.252.25 hours.

X∼N(80,100)X \sim N(80, 100). Random samples of size 2525 are taken. Complete Xˉ∼N(μXˉ,σXˉ2)\bar{X} \sim N(\mu_{\bar{X}}, \sigma^2_{\bar{X}}) by giving the mean and the variance of Xˉ\bar{X}.

Common mistake

Two places students divide wrongly

  • SD(Xˉ)=σn\mathrm{SD}(\bar{X}) = \frac{\sigma}{\sqrt{n}}, not σn\frac{\sigma}{n}.
  • Var(Xˉ)=σ2n\mathrm{Var}(\bar{X}) = \frac{\sigma^2}{n}, not σ2n\frac{\sigma^2}{\sqrt{n}}.

Check by squaring: (σn)2=σ2n\left(\frac{\sigma}{\sqrt{n}}\right)^2 = \frac{\sigma^2}{n}.

A researcher increases her sample size from 5050 to 200200. What happens to the standard deviation of the sample mean?

Choose one answer

Summary

  • E(Xˉ)=μE(\bar{X}) = \mu: sample means are centred on the population mean.

  • SD(Xˉ)=σn\mathrm{SD}(\bar{X}) = \frac{\sigma}{\sqrt{n}} and Var(Xˉ)=σ2n\mathrm{Var}(\bar{X}) = \frac{\sigma^2}{n}: bigger samples give less variable means.

  • To halve the standard deviation of Xˉ\bar{X}, take four times the sample.

  • If X∼N(μ,σ2)X \sim N(\mu, \sigma^2), then Xˉ∼N(μ,σ2n)\bar{X} \sim N\left(\mu, \frac{\sigma^2}{n}\right) exactly, for any nn.

  • In N(μ,σ2)N(\mu, \sigma^2) notation the second number is a variance; calculators want the standard deviation.

Finished the lesson?

Try the 3 quick checks above and the lesson completes itself, or mark it complete now.

Lesson 2 of 80% read

Save your progress with a free account