Hypothesis testing sits in Paper 3 Section A and spans both years: the language of testing and the binomial proportion test at AS, then the correlation test using the PMCC and the test for a Normal mean at A2. It's a compact strand — typically around 12 of the 300 marks — but a reliable one, since most series carry at least one substantial test question.
The probability underneath is usually the easy part. What separates marks lost from marks kept is the write-up: stating hypotheses correctly, choosing the right tail, and finishing with a conclusion that actually answers the question asked. Those habits are worth mastering on their own, apart from the calculation they wrap around.
The specification statements this topic covers. AS = Year-1 content, also assessed in the standalone AS course (8MA0); A2 = full A level only. Typical share of a 300-mark series: ≈12 marks — our estimate from the 2018–2025 papers, not an official weighting.
| Ref | Spec statement | Level |
|---|---|---|
| 5.1 | Language of hypothesis testing | AS |
| 5.1 | Correlation test (PMCC) | A2 |
| 5.2 | Binomial proportion test | AS |
| 5.3 | Test for a Normal mean | A2 |
Examiners repeatedly report hypotheses written as sentences, or in terms of a sample statistic — r, x̄, or an observed proportion — rather than the population parameter. H₀ and H₁ must name p, µ or ρ with an exact value (H₀: ρ = 0, not H₀: r = 0). Fix the parameter and its symbol before writing anything else.
One-tail versus two-tail is set by the question's claim: 'increased' is one-tail, 'changed' is two-tail. For a two-tail binomial test, E(X) = np decides which side to examine first. Getting this backwards flips the critical region even when every other step is right.
The PMCC test always concludes something about ρ, using r from a calculator or given data. Examiners note candidates forgetting |r| can't exceed 1, or comparing r to a critical value without weighing its sign — r = ±1 means the points are exactly collinear, nothing more.
Across Statistics, examiners' single most repeated comment is that a conclusion must be given in the question's own context, not just 'reject H₀'. It should read as a sentence about journey times or defect rates, in evidence language rather than a claim of certainty.
Writing 'the die is fair' or H₀: r = 0 instead of naming the population parameter. Hypotheses are claims about the whole population, so pick p, µ or ρ and give it an exact value before anything else.
Stopping at a comparison of values, or overclaiming with 'this proves…', instead of a sentence a non-statistician could follow. Finish with evidence language plus the scenario's own terms — what this actually says about journey times or defect rates.
Using the spread of a single observation rather than the sample mean, so the test statistic comes out far too small. A mean of n values varies less than one observation does — the model must reflect that with σ/√n, not σ, in the denominator.
Reading the cumulative table where a point probability is wanted, or writing P(X ≥ k) = 1 − P(X ≤ k) instead of 1 − P(X ≤ k − 1). Before the calculator, write out in words exactly which outcomes the event contains.
We haven’t published checked questions for this topic yet — a worked sample appears here only once a question has passed every check. In the meantime you can practise in the app.
Learn the test as a fixed structure, since that — not the probability underneath — is what most marks reward: name the parameter and state H₀/H₁ correctly, decide the tail from the wording, find the test statistic and compare it with a critical value or p-value, then conclude in context. The booklet supplies the machinery — the binomial model, the (X̄ − µ)/(σ/√n) sampling result, the PMCC critical-value tables — so what needs memorising is the sequence of steps and the phrasing each one demands, not the formulae.
Calculator discipline matters as much as anywhere else: write down the probability or statistic you're computing before you compute it, and quote table values in full rather than rounding early. For binomial tests, state explicitly which probability you're finding — that's exactly where off-by-one slips creep in.
Practise all three tests side by side, since exam questions often disguise which one applies until you've read the context. Past-paper mark schemes show how little a correct but uncontextualised conclusion earns — the fastest way to build the habit of finishing with a sentence about the scenario, not just the symbols.
All 19 topics: Edexcel A level Maths topic guides. Reference: formula booklet vs memorise and grade boundaries.
Ready to practise this topic properly? Start free — no card needed.
Original questions written for the Pearson Edexcel A Level Mathematics (9MA0) specification. Not affiliated with or endorsed by Pearson.