What Is Hypothesis Testing?

Alright everyone — today we’re tackling something that sounds a bit scary but actually isn’t once you get into it: hypothesis testing.

If you’ve ever sat in class and thought, “What’s all this about nulls and alternatives?” — don’t worry, you’re not alone. Every year I see the same reaction: polite nodding followed by quiet panic.

So let’s take it slowly, talk it through properly — like we’re in a real lesson.

🔙 Previous topic:

If you haven’t yet looked at the Large Data Set, check that first — it shows how real data connects to the ideas we test statistically here.

The Big Idea — Why Hypothesis Testing Exists

Here’s the thing: in statistics, we never prove anything for certain. We just look at data and ask,

“Is what I’m seeing actually real, or could it have happened by chance?”

That’s it — the heart of hypothesis testing.

We start with a default assumption — we call it the null hypothesis, written H₀ — and we only reject it if the data gives us strong enough evidence.

So think of it as a detective process.
You don’t say someone’s guilty (or that your new claim is true) unless there’s clear evidence against the null story.

H₀ and H₁ — What They Really Mean

Alright, this part gets people.

H₀ (null hypothesis): the “nothing’s going on” assumption.
Like saying, “This coin is fair.”
H₁ (alternative hypothesis): the “something’s happening” idea.
Like, “This coin is biased.”

We collect data — say, flip the coin 100 times — and then test whether the results are so unusual that the fair-coin idea (H₀) seems unlikely.

If they are, we reject H₀ and say there’s evidence the coin’s biased.
If not, we stick with H₀.

Notice the word “evidence.”
You never say proof.
OCR literally highlights that in their mark schemes: “Students should interpret results as evidence, not proof.”

The Process — Step by Step

Let’s run through it slowly.

Step 1: Write your hypotheses.
Make sure you include both. For example:

H₀: p = 0.5 (the coin is fair)

H₁: p ≠ 0.5 (the coin is biased)

(That’s a two-tailed test — we’ll come back to that later.)

Step 2: Choose your significance level.
Usually 5%, which means you’re okay with being wrong 1 time in 20.
AQA might write this as “at the 5% level of significance.”

Step 3: Collect data and find your test statistic.
Could be a z-score, a number of successes, or whatever fits your test type.

Step 4: Compare your statistic to the critical value or region.
If your result falls into that “critical region,” it’s rare enough that we reject H₀.

Step 5: Write your conclusion — in words.
And that’s where half the marks usually are!

The Famous Significance Level (and Why It’s 5%)

So, what’s this 5% about?

Imagine you’re tossing a fair coin 100 times.
You’d expect around 50 heads.
But if you got 90? That’s weird — too extreme to be random.

That’s the idea of significance.

If the chance of getting your data (or something more extreme) is below your chosen level — say 5% — then your result is “statistically significant.”

Meaning:

“The data is unlikely if H₀ were true, so we reject H₀.”

That’s the official line OCR loves to see.

And yes — 5% is standard, but sometimes they use 1% for stricter tests.

One-Tailed vs Two-Tailed Tests

Ah, here’s one that always catches people out.

One-tailed test: you’re testing in one direction.
Example: “This new drug increases recovery rate.”
So H₁: p > something.
Two-tailed test: you’re open to either direction.
Example: “This coin is biased.” (Could be either way.)
So H₁: p ≠ something.

Edexcel’s sneaky about this — sometimes they bury the direction in the question wording.
If it says “greater than” or “higher than,” that’s your clue for one-tailed.

And remember: for two-tailed tests, split your significance level across both ends.
That’s a classic exam trap.

Interpreting Results — The Part Students Forget

Right, this is where everyone rushes.

If you get a result in the critical region, you reject H₀.
If not, you “fail to reject” it.

But in an exam, you must say what that means in context.

For example:

“There is evidence to suggest the coin is biased.”
or
“There is not enough evidence to suggest the new teaching method increases pass rates.”

I once saw an AQA question lose marks because the student just wrote “Reject H₀.”
Correct — but incomplete.
They want to see that you’ve understood what that means in words.

Common Misconceptions (That Cost Marks!)

Let’s be honest — hypothesis testing is a playground for easy-to-miss mistakes.

🚫 Mistake 1: Writing “proves” instead of “suggests.”
✅ Say “there’s evidence.”

🚫 Mistake 2: Swapping H₀ and H₁.
✅ Always make H₀ the “no change” idea.

🚫 Mistake 3: Mixing up tails.
✅ Check the question for direction words like “more than” or “less than.”

🚫 Mistake 4: Forgetting context.
✅ Always finish with a real-world sentence.

OCR’s examiner reports are full of lines like,

“Students correctly performed calculations but failed to relate their conclusion to context.”
Translation: maths done, marks gone.

Why Hypothesis Testing Actually Matters

This isn’t just for exam marks. It’s how real research works.

Psychologists use it to test therapies.
Economists use it to check if policies affect inflation.
Biologists use it to see if drugs actually work.
Criminologists — remember that one? — use it to see if new laws reduce crime.

It’s the backbone of scientific reasoning.
You’re basically asking, “Is my evidence strong enough to doubt the default story?”

That’s such a powerful way to think.

A Quick Anecdote (Because This Always Happens)

I remember once giving my class a hypothesis test question about coin tossing.
One student flipped 10 heads in a row and shouted, “Sir, it’s definitely biased!”

We did the maths — p-value was about 0.001.
So yes, statistically significant.
Then another student grabbed the coin, flipped it 20 times, and got mixed results.

Suddenly the “biased coin” looked… not so biased.

And that, in a nutshell, is why we rely on proper sample sizes and maths, not luck or gut feeling.

Hypothesis Testing in the Exam (and How to Nail It)

Alright, let’s sum it up teacher-style:

1️⃣ Write both hypotheses — in words and symbols.
2️⃣ Pick your significance level — usually 5%.
3️⃣ Show your method — critical region or test statistic.
4️⃣ Compare your result properly.
5️⃣ Conclude in context.

That’s literally what the mark scheme wants.

Oh — and make sure you state your conclusion clearly.
No maybes, no “I think.”
Just: “There is evidence that…” or “There is insufficient evidence that…”

🧭 Next topic:

See how hypothesis statistical tests are applied in criminology. Discover how it is used to identify patterns and test theories.

Final Reflection

If you strip away all the symbols, hypothesis testing is just a fancy way of saying:

“Here’s my claim — does the data back it up?”

It’s not about memorising steps; it’s about reasoning with evidence.
And once you get that, it stops being “a stats topic” and becomes a life skill.

I tell my students every year:

“You’ll forget z-scores, but you’ll never forget how to question an assumption.”

That’s the real win.

Learn to Think Like a Researcher

Start your revision for A-Level Maths today with our A Level Maths crash course, where we break down tricky topics like hypothesis testing, regression, and probability into simple, real-world explanations.

Learn not just how to calculate — but how to think statistically, the way examiners (and scientists) want you to.

About the Author

S. Mahandru is Head of Maths at Exam.tips and has more than 15 years of experience in simplifying difficult subjects such as pure maths, mechanics and statistics. He gives worked examples, clear explanations and strategies to make students succeed.

What Is Hypothesis Testing?