← Back to subjects
⭐ 0
Edexcel A-level Mathematics (9MA0) Β· Statistical hypothesis testing
Mini-Lesson

Statistical hypothesis testing

This mini-lesson covers Topic 5 of Edexcel A-level Mathematics (9MA0): the language of testing β€” null and alternative hypotheses, significance level, test statistic, 1-tail and 2-tail tests, critical value, critical region, acceptance region and p-value β€” then three tests you must be able to carry out: a test for a binomial proportion p, a test for the mean of a normal distribution, and a test for a correlation coefficient ρ.

One-tailed, 5% 5% all 5% sits in one tail Two-tailed, 5% 2.5% 2.5% the 5% is split between two tails
The critical region (red) is the set of test-statistic values for which you reject Hβ‚€.

Work through each screen, answer the questions as you go (some are wordy, some are calculations) and collect ⭐ stars. Throughout, take g = 9.8 m sβˆ’2 unless a question says otherwise. Press Start when you are ready.

Testing Β· the language

Hypotheses, significance and the critical region

  • Hβ‚€, the null hypothesis β€” always a statement of equality about a population parameter (p, ΞΌ or ρ). It is what you assume is true while you calculate.
  • H₁, the alternative hypothesis β€” what you are testing for. Use > or < for a one-tailed test and β‰  for a two-tailed test.
  • Significance level (Ξ±) β€” the probability of incorrectly rejecting Hβ‚€. Typically 10%, 5% or 1%.
  • Test statistic β€” the value calculated from the sample.
  • Critical region β€” the values of the test statistic that lead you to reject Hβ‚€. Everything else is the acceptance region.
  • p-value β€” the probability, assuming Hβ‚€ is true, of getting the observed result or one more extreme. Reject Hβ‚€ if p-value < Ξ±.

Two-tailed rule: at the 5% level, each tail gets 2.5%. Comparing a one-tailed p-value with 0.025 (or doubling it and comparing with 0.05) is the standard method.

Never say "Hβ‚€ is true". The correct wording is: "there is sufficient / insufficient evidence to reject Hβ‚€ at the 5% level", followed by a sentence in context.

Quick check

Stating Hβ‚€

?In a test on a binomial proportion, how must the null hypothesis be stated?
Sort it

Hβ‚€, H₁, or a conclusion?

Tap a statement, then tap where it belongs.

0️⃣ Null hypothesis Hβ‚€

1️⃣ Alternative hypothesis H₁

✍️ Conclusion wording

Testing Β· binomial proportion

Testing a binomial proportion p

Model the count with X ~ B(n, p), assume Hβ‚€ is true, then find the probability of the observed value or more extreme.

Worked example β€” one-tailed

A coin-flip style claim: p = 0.3. In n = 20 trials there are 10 successes. Test at the 5% level whether p has increased.

Hβ‚€: p = 0.3  Β·  H₁: p > 0.3 (one-tailed)

Assume X ~ B(20, 0.3). p-value = P(X β‰₯ 10) = 1 βˆ’ P(X ≀ 9) = 1 βˆ’ 0.9520 = 0.0480.

0.0480 < 0.05, so reject Hβ‚€: there is sufficient evidence at the 5% level that the proportion has increased.

Worked example β€” two-tailed

X ~ B(10, 0.5) under Hβ‚€: p = 0.5, with H₁: p β‰  0.5, at the 5% level. Observed: 9 successes.

9 is above the expected value np = 5, so use the upper tail: P(X β‰₯ 9) = (10 + 1) Γ— 0.5¹⁰ = 11 Γ· 1024 = 0.01074.

Two-tailed: compare 0.01074 with 0.025 β€” or double it, 2 Γ— 0.01074 = 0.0215, and compare with 0.05.

0.0215 < 0.05, so reject Hβ‚€: there is evidence the probability is not 0.5.

Calculate

Binomial p-value

1Hβ‚€: p = 0.3, H₁: p > 0.3, with n = 20 trials and 10 successes. Given that P(X ≀ 9) = 0.9520 for X ~ B(20, 0.3), find the p-value P(X β‰₯ 10), to 4 decimal places.
Hint: P(X β‰₯ 10) = 1 βˆ’ P(X ≀ 9) = 1 βˆ’ 0.9520. Compare your answer with 0.05.
Calculate

Two-tailed p-value

2Two-tailed test, X ~ B(10, 0.5) under Hβ‚€, with 9 successes observed. Find the doubled p-value, 2 Γ— P(X β‰₯ 9), to 4 decimal places.
Hint: P(X β‰₯ 9) = P(X = 9) + P(X = 10) = (10 + 1) Γ— 0.5¹⁰ = 11 Γ· 1024 = 0.010742. Now double it.
Quick check

Splitting the tails

?A two-tailed test is carried out at the 5% significance level. What probability sits in each tail?
Testing Β· normal mean

Testing the mean of a normal distribution

If X ~ N(ΞΌ, σ²), then the sample mean of n observations satisfies:

XΜ„ ~ N(ΞΌ, σ² Γ· n)so z = (xΜ„ βˆ’ ΞΌ) Γ· (Οƒ Γ· √n) and Z ~ N(0, 1Β²)

Note the variance is divided by n: bigger samples give a more reliable mean. Hypotheses are stated in terms of the population mean ΞΌ.

Worked example

X ~ N(ΞΌ, 5Β²). A sample of n = 25 gives a sample mean of xΜ„ = 52. Test Hβ‚€: ΞΌ = 50 against H₁: ΞΌ > 50 at the 5% level.

Standard error = Οƒ Γ· √n = 5 Γ· √25 = 5 Γ· 5 = 1.

z = (52 βˆ’ 50) Γ· 1 = 2. p-value = P(Z > 2) = 1 βˆ’ 0.9772 = 0.0228.

0.0228 < 0.05 β†’ reject Hβ‚€. There is evidence at the 5% level that the mean has increased.

Critical value method: for a one-tailed 5% test the critical value is z = 1.6449; since 2 > 1.6449, reject Hβ‚€ β€” the same conclusion.

Calculate

Test statistic

3X ~ N(ΞΌ, 5Β²). A sample of 25 gives xΜ„ = 52. Testing Hβ‚€: ΞΌ = 50, find the test statistic z.
Hint: z = (xΜ„ βˆ’ ΞΌ) Γ· (Οƒ Γ· √n) = (52 βˆ’ 50) Γ· (5 Γ· √25) = 2 Γ· 1.
Calculate

p-value for the mean

4Using z = 2 from the last question, find the p-value P(Z > 2) for this one-tailed test, to 4 decimal places.
Hint: P(Z > 2) = 1 βˆ’ Ξ¦(2) = 1 βˆ’ 0.9772. Compare with 0.05.
Calculate

Critical value of xΜ„

5For the same test but two-tailed at 5% (H₁: ΞΌ β‰  50), the critical z-values are Β±1.96 and Οƒ Γ· √n = 1. Find the upper critical value of xΜ„.
Hint: xΜ„ = ΞΌ + z Γ— (Οƒ Γ· √n) = 50 + 1.96 Γ— 1. (The lower critical value is 50 βˆ’ 1.96 = 48.04.)
Match it

Term β†’ definition

Tap an item on the left, then its partner on the right.

Definition
Term
Testing Β· correlation

Testing a correlation coefficient ρ

The product moment correlation coefficient r measures how close the sample's points lie to a straight line, with |r| ≀ 1. We use it to test the population correlation coefficient, ρ (rho).

  • Hβ‚€: ρ = 0 β€” always. ("There is no correlation in the population.")
  • H₁: ρ > 0 or ρ < 0 (one-tailed), or ρ β‰  0 (two-tailed).
  • You are given a table of critical values (or a p-value). Compare |r| with the critical value for that sample size and significance level.
Worked example

A sample of n = 12 pairs gives r = 0.62. Test at the 5% level whether there is positive correlation.

Hβ‚€: ρ = 0  Β·  H₁: ρ > 0 (one-tailed). The critical value from the table is 0.4973.

r = 0.62 > 0.4973, so the result is in the critical region β†’ reject Hβ‚€.

In context: there is sufficient evidence at the 5% level of a positive correlation between the two variables.

You are not asked to calculate r by hand β€” a calculator (or the question) supplies it. What is examined is the hypotheses, the comparison and the conclusion in context.

Quick check

Hypotheses for ρ

?What is the correct null hypothesis when testing a correlation coefficient?
Quick check

What a p-value is

?Which of these correctly describes a p-value?
Testing Β· critical regions

Finding a binomial critical region

Instead of finding a p-value you can find the critical region in advance: the smallest set of extreme values whose total probability is at most the significance level.

Worked example β€” X ~ B(20, 0.3), H₁: p > 0.3, at 5%

Work down from the top until the tail probability exceeds 5%:

P(X β‰₯ 10) = 1 βˆ’ P(X ≀ 9) = 1 βˆ’ 0.9520 = 0.0480 ≀ 0.05 βœ”

P(X β‰₯ 9) = 1 βˆ’ P(X ≀ 8) = 1 βˆ’ 0.8867 = 0.1133 > 0.05 ✘

So the critical region is X β‰₯ 10, and the actual significance level of the test is 0.0480, i.e. 4.80% β€” not exactly 5%, because X is discrete.

Actual significance level = the true probability of the test statistic falling in the critical region when Hβ‚€ is true. For a discrete distribution it is almost never exactly Ξ± β€” this is a favourite exam question, and the answer is a probability, not "5%".

Quick check

Actual significance level

?For the test above, the critical region is X β‰₯ 10 and P(X β‰₯ 10) = 0.0480. What is the actual significance level of the test?
Recap

The big ideas to know

Hβ‚€: always an equality about a population parameter (p, ΞΌ or ρ = 0) Β· H₁: >, < (1-tail) or β‰  (2-tail)

Significance level: the probability of incorrectly rejecting Hβ‚€; two-tailed β‡’ split it, 2.5% per tail

Binomial test: assume X ~ B(n, pβ‚€), find P(X β‰₯ x) or P(X ≀ x), compare with Ξ±

Normal mean test: XΜ„ ~ N(ΞΌ, σ²/n), so z = (xΜ„ βˆ’ ΞΌ) Γ· (Οƒ Γ· √n)

Correlation test: Hβ‚€: ρ = 0; compare |r| with the given critical value

Conclusion: β€œsufficient / insufficient evidence to reject Hβ‚€ at the 5% level” β€” then say what it means in context

That is the whole of 9MA0 Topic 5 β€” Statistical hypothesis testing. Press Finish to see your score.

πŸ†

Mini-lesson complete!

⭐⭐⭐

You have worked through Statistical hypothesis testing for Edexcel A-level Mathematics (9MA0). πŸŽ‰

Your stars: 0 / 0

Next: test yourself in the Evaluate stage Confidence Quiz, then lock it in with Verify.

πŸ“£ Smashed it? Share your score

Challenge a mate to beat your stars, or show a parent how you got on.

β†’ Back to all subjects