Hypothesis test for the mean of a Normal distribution

Hypothesis test for the mean of a Normal distribution

Spec O3 · AQA A-Level Maths: Revision notes

Studying this course? Keep your progress in one place.

Key facts

  • H0H_0 states the claim with an equals sign, about the population mean μ\mu, never xˉ\bar x.
  • 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); the standard error is σn\frac{\sigma}{\sqrt n}.
  • Reject H0H_0 if the pp-value is below the significance level, or the sample mean is in the critical region.
  • For a two-tailed test, split the significance level between the tails.
  • Conclude in two parts: the decision, and what it means in context. Never say H0H_0 is proved.

Hypotheses and significance

Whether sample data give enough evidence against a claim about a population mean.

A hypothesis test decides whether sample data give enough evidence against a claim. The null hypothesis H0H_0 states the claim, such as H0:μ=250H_0:\mu=250. The alternative hypothesis H1H_1 says how the parameter differs: μ<250\mu<250 or μ>250\mu>250 (one-tailed) or μ≠250\mu\ne250 (two-tailed). Hypotheses are about μ\mu, never the sample mean.

The significance level is the probability, assuming H0H_0, below which we reject H0H_0 (commonly 5%5\% or 1%1\%). The critical region is the set of results that lead to rejecting H0H_0, and the p-value is the probability, assuming H0H_0, of a result at least as extreme as observed.

−4−3−2−112340.10.20.30.4xyz = 1.645standard Normal
Standard Normal curve with the 5% upper-tail critical region (z > 1.645) shaded.

A manufacturer claims bottles hold 250250 ml on average. A test checks whether they hold less. Which hypotheses are correct?

The sample mean

Xˉ∼N(μ,σ2n)\bar X\sim N\left(\mu,\frac{\sigma^2}{n}\right): the sample mean varies less than a single observation.

If X∼N(μ,σ2)X\sim N(\mu,\sigma^2) and a random sample of size nn is taken, the sample mean has the distribution Xˉ∼N(μ,σ2n)\bar X\sim N\left(\mu,\frac{\sigma^2}{n}\right).

Its standard deviation, σn\frac{\sigma}{\sqrt n}, is the standard error. It falls as nn rises. The test needs σ\sigma to be known, given or assumed.

2402422442462482502522542562582600.20.40.60.811.21.4xyX̄ ~ N(250, 36/n)
The distribution of Xˉ\bar X for μ=250\mu=250, σ=6\sigma=6: a larger sample nn makes it narrower. Try the slider.
  • Xˉ\bar XN(μ, σ2n)N\left(\mu,\ \frac{\sigma^2}{n}\right)

Worked example

Under H0:μ=250H_0:\mu=250, σ=6\sigma=6 and n=36n=36. State the distribution of Xˉ\bar X.

σ=10\sigma=10 and n=25n=25. What is the standard error?

One-tailed test

State the hypotheses, write the distribution of Xˉ\bar X under H0H_0, find the pp-value, compare, and conclude in context.

  1. State H0H_0, H1H_1 and the significance level.
  2. Write the distribution of Xˉ\bar X assuming H0H_0.
  3. Find the pp-value, or the critical value.
  4. Reject H0H_0 if the pp-value is below the significance level, or xˉ\bar x lies in the critical region.
  5. Write the conclusion in context.

A zz-value works too: for a 5%5\% one-tailed test, reject if z>1.645z>1.645 (upper) or z<−1.645z<-1.645 (lower).

  1. 1

    Hypotheses

    state H₀, H₁ and the level

  2. 2

    Distribution

    of X̄ assuming H₀

  3. 3

    p-value

    or the critical value

  4. 4

    Compare

    reject H₀ or not

  5. 5

    Conclude

    in context

A one-tailed test in five steps

Worked example

H0:μ=48H_0:\mu=48, H1:μ>48H_1:\mu>48 at the 5%5\% level, with σ=5\sigma=5, n=40n=40 and xˉ=49.6\bar x=49.6. Carry out the test.

In a 5%5\% upper-tail test you find z=1.5z=1.5. What is the decision?

Two-tailed tests

The critical region is split between both tails: 2.5%2.5\% in each at the 5%5\% level, so z=±1.96z=\pm1.96.

If H1:μ≠μ0H_1:\mu\ne\mu_0, the critical region is in both tails. At the 5%5\% level put 2.5%2.5\% in each tail, so the critical values of zz are ±1.96\pm1.96. Alternatively double the one-tail probability to get the pp-value and compare it with 5%5\%. This is a two-tailed test.

−4−3−2−112340.10.20.30.4xyz = −1.96z = 1.96standard Normal
A 5% two-tailed test: 2.5% in each tail, beyond z = ±1.96.

Worked example

H0:μ=30H_0:\mu=30, H1:μ≠30H_1:\mu\ne30, σ=4\sigma=4, n=16n=16, xˉ=31.8\bar x=31.8. Test at the 5%5\% level.

For a two-tailed test at the 1%1\% level, how much probability goes in each tail?

Conclusions and assumptions

Give the decision about H0H_0 and what it means in context; a larger sample gives stronger evidence.

Always write a two-part conclusion. If the pp-value is less than the significance level, reject H0H_0: 'there is sufficient evidence at the 5%5\% level that the mean journey time is longer than 4848 minutes'. Otherwise: 'there is insufficient evidence'. Never say H0H_0 is proved true or false.

The same difference in sample mean gives stronger evidence with a larger nn: a mean of 49.649.6 gives p=0.0215p=0.0215 with n=40n=40 but p=0.156p=0.156 with n=10n=10. The test assumes the sample is random, the population is Normal, and σ\sigma is known, given or assumed.

00.020.040.060.080.10.120.140.16n = 10n = 40Sample size np-value
The same sample mean gives a smaller p-value with a larger sample.

p=0.08p=0.08 at the 5%5\% level. Which conclusion is correct?

Try an exam question

The mass of a box of cereal is Normal with standard deviation 88 g. A claim says the mean is 500500 g. A random sample of 6464 boxes has mean 498.4498.4 g. Test, at the 5%5\% significance level, whether the mean mass is less than 500500 g.

[4 marks]

That's the notes covered.

Carry on to the next subtopic.

Try a question

Assuming the mean is 250250 g, the distribution of the sample mean Xˉ\bar X of 3636 bags is

1 mark, marked against the mark scheme

Revision notes