Step-by-step worked examples and graded practice questions on hypothesis testing with the binomial distribution — null and alternative hypotheses, significance levels, one-tailed tests using probabilities, critical values and critical regions, actual significance levels, two-tailed tests, and writing conclusions in context. Written to the Edexcel Statistics Year 1 specification and equally suitable for AQA and OCR A.
📚 Statistics Year 1 (AS)✅ 15 Practice Questions🔍 7 Worked Examples⚠️ Common Mistakes
Struggling with hypothesis testing?
Alamin diagnoses exactly which skills are missing and builds a structured plan to fix them — backed by AI-powered practice between sessions.
A hypothesis test uses a sample to decide whether there is enough evidence to doubt a claim about a whole population. At AS level the claim is always about a probability or proportion, \(p\), and the number of successes in the sample is modelled by the binomial distribution from Statistical Distributions (Binomial). The idea is the same one met in GCSE probability when deciding whether a die is biased: a result that would be very unlikely for a fair die is evidence that it is not fair.
Four skills make up this topic:
The language of hypothesis testing — null and alternative hypotheses, the test statistic, the significance level, and one-tailed and two-tailed tests
Testing with a probability — finding the probability of a result at least as extreme as the one observed, and comparing it with the significance level
Critical regions — the critical value, the critical region and acceptance region, and the actual significance level
Two-tailed tests — splitting the significance level between the two tails, and writing every conclusion in context
The notation follows the binomial topic: \(X \sim \text{B}(n, p)\), with cumulative probabilities \(\text{P}(X \leq x)\) from a calculator or tables. Probabilities that are not exact are given to 4 decimal places, and full calculator values are kept until the final answer.
The language of hypothesis testing
Every test starts from a claim about the population parameter \(p\), the probability of a success on each trial. The sample is used to make an inference about that population value.
The null hypothesis, \(\text{H}_0\), is the claim that is assumed to be true, for example \(\text{H}_0: p = 0.3\). It always gives a single value of \(p\), so that the distribution of the test statistic is known.
The alternative hypothesis, \(\text{H}_1\), says how \(p\) is believed to differ: \(p > 0.3\) or \(p < 0.3\) for a one-tailed test, or \(p \neq 0.3\) for a two-tailed test, when the direction of the change is not known.
The test statistic is the quantity calculated from the sample, here \(X\), the number of successes. If \(\text{H}_0\) is true, \(X \sim \text{B}(n, p)\) with the value of \(p\) from \(\text{H}_0\).
The significance level, such as 5%, sets how unlikely a result must be before \(\text{H}_0\) is rejected. It is the probability of incorrectly rejecting \(\text{H}_0\), that is, rejecting \(\text{H}_0\) when it is in fact true, so a test can give the wrong conclusion even when it is carried out correctly. For a binomial test, the exact value of this probability is the actual significance level, met in the section on critical regions.
Hypotheses are always written in terms of the population parameter \(p\), never in terms of the sample result. Define \(p\) in words, in the context of the question.
Worked Example 1
For each situation, define the parameter \(p\) and write down suitable null and alternative hypotheses. State whether the test is one-tailed or two-tailed.(a) A coin is suspected of landing heads more often than tails.(b) A bus company claims that 80% of its buses arrive on time. A passenger believes that the proportion is lower.(c) In the past, 35% of the students at a school cycled to school. A governor thinks that this proportion has changed.
1
(a) Let \(p\) be the probability that the coin lands heads. A fair coin has \(p = 0.5\), and the suspicion is that \(p\) is larger: \(\text{H}_0: p = 0.5\), \(\text{H}_1: p > 0.5\). This is one-tailed
2
(b) Let \(p\) be the proportion of buses that arrive on time. The claim is \(p = 0.8\), and the passenger believes it is smaller: \(\text{H}_0: p = 0.8\), \(\text{H}_1: p < 0.8\). This is one-tailed
3
(c) Let \(p\) be the proportion of students who cycle to school. "Changed" gives no direction, so \(\text{H}_0: p = 0.35\), \(\text{H}_1: p \neq 0.35\). This is two-tailed
Answer(a) \(\text{H}_0: p = 0.5\), \(\text{H}_1: p > 0.5\); one-tailed(b) \(\text{H}_0: p = 0.8\), \(\text{H}_1: p < 0.8\); one-tailed(c) \(\text{H}_0: p = 0.35\), \(\text{H}_1: p \neq 0.35\); two-tailed
Carrying out a one-tailed test
One way to carry out a test is to find the probability, assuming \(\text{H}_0\) is true, of a result at least as extreme as the one observed. This probability is called the p-value. If it is less than or equal to the significance level, the result is significant and \(\text{H}_0\) is rejected.
Define \(p\) and write down \(\text{H}_0\) and \(\text{H}_1\).
State the distribution of the test statistic assuming \(\text{H}_0\) is true, such as \(X \sim \text{B}(20, 0.8)\).
Find the probability of the observed value or anything more extreme in the direction of \(\text{H}_1\): \(\text{P}(X \geq x)\) for \(\text{H}_1: p > \ldots\), or \(\text{P}(X \leq x)\) for \(\text{H}_1: p < \ldots\).
Compare this probability with the significance level, and decide whether to reject \(\text{H}_0\).
Write a conclusion in the context of the question.
Use the probability of the whole tail, never \(\text{P}(X = x)\) on its own: with a large sample, every single value has a small probability. Conclusions should not overstate the result. Write "there is evidence that…" or "there is insufficient evidence that…", because a test never proves that a hypothesis is true or false.
Worked Example 2
A die is suspected of being biased towards six. It is rolled 30 times and lands on six 10 times. Test, at the 5% level of significance, whether there is evidence that the die is biased towards six.
1
Let \(p\) be the probability that the die lands on six. \(\text{H}_0: p = \dfrac{1}{6}\), \(\text{H}_1: p > \dfrac{1}{6}\)
2
Let \(X\) be the number of sixes in 30 rolls. Assuming \(\text{H}_0\) is true, \(X \sim \text{B}\left(30, \dfrac{1}{6}\right)\)
3
A fair die would give about \(30 \times \dfrac{1}{6} = 5\) sixes, and 10 is more than this, so find the upper tail: \(\text{P}(X \geq 10) = 1 - \text{P}(X \leq 9) = 1 - 0.980293\ldots = 0.019706\ldots\)
4
\(0.0197 < 0.05\), so the result is significant at the 5% level. Reject \(\text{H}_0\)
5
There is evidence, at the 5% level, that the die is biased towards six
Answer\(\text{P}(X \geq 10) = 0.0197 < 0.05\), so reject \(\text{H}_0\): there is evidence that the die is biased towards six
The distribution of \(X \sim \text{B}\left(30, \dfrac{1}{6}\right)\), assuming \(\text{H}_0\) is true. The amber bars, \(X = 10\) to 30, make up the p-value \(\text{P}(X \geq 10) = 0.0197\) in Worked Example 2. The bars above \(X = 12\) are too small to see.
Worked Example 3
The bus company in Worked Example 1(b) claims that 80% of its buses arrive on time. The passenger records a random sample of 20 buses, and 13 of them arrive on time. Test the passenger's belief at the 5% level of significance.
1
Let \(p\) be the proportion of buses that arrive on time. \(\text{H}_0: p = 0.8\), \(\text{H}_1: p < 0.8\)
2
Let \(X\) be the number of buses in the sample that arrive on time. Assuming \(\text{H}_0\) is true, \(X \sim \text{B}(20, 0.8)\)
3
\(\text{H}_1\) is \(p < 0.8\), so find the lower tail: \(\text{P}(X \leq 13) = 0.086692\ldots\)
4
\(0.0867 > 0.05\), so the result is not significant at the 5% level. Do not reject \(\text{H}_0\)
5
There is insufficient evidence, at the 5% level, that fewer than 80% of the buses arrive on time. This does not prove the company's claim: the sample simply does not give strong enough evidence against it
Answer\(\text{P}(X \leq 13) = 0.0867 > 0.05\), so do not reject \(\text{H}_0\): there is insufficient evidence that fewer than 80% of buses arrive on time
Critical regions
Instead of working out a probability for the observed value, you can first find every value of the test statistic that would lead to \(\text{H}_0\) being rejected.
The critical region is the set of values that would lead to rejecting \(\text{H}_0\). The other values form the acceptance region.
The critical value is the first value inside the critical region, on the boundary with the acceptance region.
The critical region is chosen so that its probability, assuming \(\text{H}_0\) is true, is as close as possible to the significance level without going over it.
Because \(X\) takes only whole-number values, that probability is usually smaller than the significance level. It is called the actual significance level, and it is the probability of incorrectly rejecting \(\text{H}_0\).
To find a critical region, write down the cumulative probabilities either side of the boundary, so that it is clear the value you choose is in the region and the next value is not.
Worked Example 4
In the past, 30% of the visitors to a website bought something. After the website is redesigned, the owner takes a random sample of 20 visitors to test whether the proportion who buy something has increased, using a 5% level of significance.(a) Find the critical region for the test.(b) State the actual significance level of the test.(c) In the sample, 9 visitors buy something. State the conclusion of the test.
1
(a) Let \(p\) be the proportion of visitors who buy something. \(\text{H}_0: p = 0.3\), \(\text{H}_1: p > 0.3\). Assuming \(\text{H}_0\) is true, \(X \sim \text{B}(20, 0.3)\)
2
The critical region is in the upper tail. We need the smallest \(c\) with \(\text{P}(X \geq c) \leq 0.05\)
3
\(\text{P}(X \geq 9) = 1 - \text{P}(X \leq 8) = 1 - 0.886668\ldots = 0.113331\ldots\), which is more than 0.05
4
\(\text{P}(X \geq 10) = 1 - \text{P}(X \leq 9) = 1 - 0.952038\ldots = 0.047961\ldots\), which is less than 0.05. So the critical value is 10, and the critical region is \(X \geq 10\)
5
(b) The actual significance level is \(\text{P}(X \geq 10) = 0.047961\ldots\)
6
(c) 9 is not in the critical region, so do not reject \(\text{H}_0\). There is insufficient evidence, at the 5% level, that the proportion of visitors who buy something has increased
Answer(a) \(X \geq 10\)(b) 0.0480(c) Do not reject \(\text{H}_0\): insufficient evidence that the proportion has increased
The distribution of \(X \sim \text{B}(20, 0.3)\), assuming \(\text{H}_0\) is true. The amber bars are the critical region \(X \geq 10\) from Worked Example 4, with total probability 0.0480. The teal bars, \(X = 0\) to 9, are the acceptance region.
Worked Example 5
A college claims that 40% of its students walk to college. A student believes that the proportion is lower, and asks a random sample of 25 students.(a) Using a 5% level of significance, find the critical region for a test of the student's belief.(b) State the probability of incorrectly rejecting \(\text{H}_0\) with this test.(c) In the sample, 5 students walk to college. Complete the test.
1
(a) Let \(p\) be the proportion of students who walk to college. \(\text{H}_0: p = 0.4\), \(\text{H}_1: p < 0.4\). Assuming \(\text{H}_0\) is true, \(X \sim \text{B}(25, 0.4)\)
2
The critical region is in the lower tail. We need the largest \(c\) with \(\text{P}(X \leq c) \leq 0.05\)
3
\(\text{P}(X \leq 5) = 0.029362\ldots\), which is less than 0.05, and \(\text{P}(X \leq 6) = 0.073565\ldots\), which is more than 0.05. So the critical region is \(X \leq 5\)
4
(b) The probability of incorrectly rejecting \(\text{H}_0\) is the actual significance level, \(\text{P}(X \leq 5) = 0.029362\ldots\)
5
(c) 5 is in the critical region, so reject \(\text{H}_0\). There is evidence, at the 5% level, that fewer than 40% of the students walk to college
Answer(a) \(X \leq 5\)(b) 0.0294(c) Reject \(\text{H}_0\): there is evidence that fewer than 40% of students walk to college
Two-tailed tests
When \(\text{H}_1\) is \(p \neq \ldots\), a result that is unusually high or unusually low counts as evidence against \(\text{H}_0\). The significance level is split equally between the two tails, so a 5% two-tailed test has at most 2.5% in each tail.
To find a critical region, find a lower tail and an upper tail, each with probability as close as possible to half the significance level without going over it. The actual significance level is the sum of the two tail probabilities.
To test an observed value, first decide which tail it is in. If \(\text{H}_0\) is true, you would expect about \(n \times p\) successes: a result above this is tested in the upper tail, and a result below it in the lower tail. Compare the tail probability with half the significance level.
Worked Example 6
A spinner is claimed to land on red with probability 0.4. To test whether this is wrong, the spinner is spun 20 times and the number of reds, \(X\), is recorded.(a) Using a 5% level of significance, find the critical region for the test. The probability in each tail should be as close as possible to 0.025 without going over it.(b) State the actual significance level of the test.
1
(a) Let \(p\) be the probability that the spinner lands on red. \(\text{H}_0: p = 0.4\), \(\text{H}_1: p \neq 0.4\). Assuming \(\text{H}_0\) is true, \(X \sim \text{B}(20, 0.4)\)
2
Lower tail: \(\text{P}(X \leq 3) = 0.015961\ldots\), which is less than 0.025, and \(\text{P}(X \leq 4) = 0.050951\ldots\), which is more. So the lower tail is \(X \leq 3\)
3
Upper tail: \(\text{P}(X \geq 12) = 1 - \text{P}(X \leq 11) = 1 - 0.943473\ldots = 0.056526\ldots\), which is more than 0.025
4
\(\text{P}(X \geq 13) = 1 - \text{P}(X \leq 12) = 1 - 0.978971\ldots = 0.021028\ldots\), which is less than 0.025. So the upper tail is \(X \geq 13\)
The distribution of \(X \sim \text{B}(20, 0.4)\), assuming \(\text{H}_0\) is true. The amber bars are the two tails of the critical region in Worked Example 6: \(X \leq 3\), with probability 0.0160, and \(X \geq 13\), with probability 0.0210. The bars above \(X = 15\) are too small to see.
Worked Example 7
A survey claims that 40% of the adults in a town own a dog. A vet thinks that this figure is wrong. In a random sample of 30 adults from the town, 18 own a dog. Test, at the 5% level of significance, whether the vet is correct.
1
Let \(p\) be the proportion of adults in the town who own a dog. \(\text{H}_0: p = 0.4\), \(\text{H}_1: p \neq 0.4\). Assuming \(\text{H}_0\) is true, \(X \sim \text{B}(30, 0.4)\)
2
If \(\text{H}_0\) is true, you would expect about \(30 \times 0.4 = 12\) dog owners. 18 is above this, so use the upper tail
The test is two-tailed, so compare with half of 5%: \(0.0212 < 0.025\). The result is significant, so reject \(\text{H}_0\). (Equivalently, the two-tailed p-value is \(2 \times 0.021239\ldots = 0.042479\ldots\), which is less than 0.05)
5
There is evidence, at the 5% level, that the proportion of adults in the town who own a dog is not 40%
Answer\(\text{P}(X \geq 18) = 0.0212 < 0.025\), so reject \(\text{H}_0\): there is evidence that the proportion of adults who own a dog is not 40%
📋 Exam board note
What this topic does and does not include
In the Edexcel specification, this topic is the language of hypothesis testing developed through a binomial model, and hypothesis tests for the proportion \(p\) of a binomial distribution, with conclusions interpreted in context. Students should appreciate that the significance level is the probability of incorrectly rejecting the null hypothesis; a formal understanding of Type I errors is not expected there, so it is not taught on this page. If you are not studying Edexcel, check your own board's specification for exactly which terms and types of test you need at AS. Hypothesis tests for a correlation coefficient and for the mean of a normal distribution are full A-level content for Statistics Year 2, whose topic pages are not yet published.
Exam tips
💡 Tip 1
Define p and write both hypotheses
Say in words what \(p\) is, in context, then write \(\text{H}_0\) and \(\text{H}_1\) in terms of \(p\). Both usually earn a mark, and \(\text{H}_1\) tells you which tail to use.
💡 Tip 2
State the distribution under the null hypothesis
Write "Assuming \(\text{H}_0\) is true, \(X \sim \text{B}(n, p)\)" with the numbers in, and say what \(X\) counts. Every probability in the test comes from this distribution.
💡 Tip 3
Compare like with like
Compare a tail probability with the significance level, or compare the observed value with the critical region. In a two-tailed test, compare a single tail with half the significance level.
💡 Tip 4
Show the probabilities either side of a critical value
When finding a critical region, write down the probability for the critical value and for the next value outside the region, such as \(\text{P}(X \geq 10) = 0.0480\) and \(\text{P}(X \geq 9) = 0.1133\). This shows why the boundary is where it is.
💡 Tip 5
Conclude in context, without overclaiming
Give two parts: the decision ("reject \(\text{H}_0\)" or "do not reject \(\text{H}_0\)") and what it means in context ("there is evidence that the die is biased towards six"). Avoid "proves" and "accept \(\text{H}_0\)".
Common mistakes
Common Mistake 1
Writing hypotheses about the sample
"\(\text{H}_0: X = 10\)" or a hypothesis about the sample proportion is wrong. Hypotheses are about the population parameter: \(\text{H}_0: p = 0.3\).
Common Mistake 2
Using \(\text{P}(X = x)\) instead of the tail
The evidence is the probability of the observed value or anything more extreme. For 10 sixes, use \(\text{P}(X \geq 10)\), not \(\text{P}(X = 10)\).
Common Mistake 3
Forgetting to halve the significance level
In a 5% two-tailed test, each tail is compared with 0.025, not 0.05. A tail probability of 0.03 is significant in a one-tailed test but not in a two-tailed test.
Common Mistake 4
Taking the wrong complement for an upper tail
\(\text{P}(X \geq 10) = 1 - \text{P}(X \leq 9)\), not \(1 - \text{P}(X \leq 10)\). The value 10 must stay in the tail.
Common Mistake 5
Overstating the conclusion
Not rejecting \(\text{H}_0\) does not show that \(\text{H}_0\) is true, and rejecting it does not prove \(\text{H}_1\). Write "there is (insufficient) evidence that…", in the context of the question.
Practice questions
Work through each question before checking the worked solution. Write hypotheses in terms of \(p\), and give probabilities to 4 decimal places.
Core Skills
Q1For each situation, define \(p\), write down suitable null and alternative hypotheses, and state whether the test is one-tailed or two-tailed.(a) A manufacturer claims that 5% of its light bulbs are faulty. A shop owner thinks that the proportion is higher.(b) A coin is to be tested to see whether it is biased.(c) In the past, 70% of patients recovered with the standard treatment. A doctor believes that a new treatment is less effective.Core Skills
Q2A coin is tossed 20 times and lands heads 15 times. Test, at the 5% level of significance, whether the coin is biased towards heads.Core Skills
Q3A test uses a sample of 15 trials, with \(\text{H}_0: p = 0.4\) and \(\text{H}_1: p < 0.4\), at the 5% level of significance. \(X\) is the number of successes.(a) Find the critical region.(b) State the actual significance level.(c) The sample gives \(x = 3\). State the conclusion of the test.Core Skills
Q4A coin is tossed 12 times to test \(\text{H}_0: p = 0.5\) against \(\text{H}_1: p \neq 0.5\), where \(p\) is the probability of a head. The critical region is \(X \leq 2\) or \(X \geq 10\), where \(X\) is the number of heads.(a) Explain why this is a two-tailed test.(b) Find the actual significance level of the test.(c) The coin lands heads 9 times. State the conclusion of the test.Core Skills
Q5A council says that 30% of the households in a town recycle their food waste. A campaigner thinks that the true proportion is lower. In a random sample of 25 households, 4 recycle their food waste. Test the campaigner's belief at the 5% level of significance.Core Skills
Exam-Style
Q6A seed company states that 20% of the seeds of one variety produce white flowers. A gardener plans to test this statement with a two-tailed test, using a random sample of 25 seeds and a 5% level of significance.(a) Find the critical region. The probability in each tail should be as close as possible to 0.025 without going over it.(b) State the actual significance level.(c) 9 of the 25 seeds produce white flowers. State the conclusion of the test.Exam-Style
Q7In the past, 15% of the visitors to a museum bought a guidebook. After a new display opens, 14 out of a random sample of 50 visitors buy a guidebook.(a) Test, at the 1% level of significance, whether the proportion of visitors who buy a guidebook has increased.(b) State, with a reason, whether your conclusion would be different at the 5% level.Exam-Style
Q8In the past, 35% of a company's customers paid by direct debit. A manager believes that this proportion has changed. In a random sample of 40 customers, 8 pay by direct debit. Test the manager's belief at the 5% level of significance.Exam-Style
Q9A drug causes side effects in 25% of patients. A researcher suspects that a new version of the drug causes side effects in a higher proportion of patients, and plans a test using a random sample of 30 patients.(a) Using a 5% level of significance, find the critical region for the test.(b) State the probability of incorrectly rejecting the null hypothesis.(c) 12 of the 30 patients have side effects. State the conclusion of the test.Exam-Style
Q10A newspaper claims that 60% of adults support a plan. A politician believes that this figure is wrong. In a random sample of 20 adults, 7 support the plan.(a) Explain why a two-tailed test is appropriate.(b) Test the politician's belief at the 10% level of significance.(c) A journalist writes that the result "proves the newspaper's figure is wrong". Comment on this statement.Exam-Style
A* Challenge
Q11A coin is tossed 20 times and lands heads 14 times.(a) Test, at the 5% level of significance, whether the coin is biased towards heads.(b) State, with a reason, the conclusion of the same one-tailed test at the 10% level.(c) A second student uses the same results to test whether the coin is biased, with \(\text{H}_1: p \neq 0.5\), at the 10% level. Carry out this test, and explain why its conclusion differs from (b).A* Challenge
Q12In the past, 10% of the items made by a machine were defective. After the machine is serviced, an engineer tests whether the proportion of defective items has changed, using a random sample of 10 items and a two-tailed test at the 5% level.(a) Show that the critical region has no lower tail, and find the upper tail.(b) Explain what this means for the engineer.(c) Find the smallest sample size \(n\) for which the lower tail of the critical region is not empty.A* Challenge
Q13A factory states that 10% of its components are faulty. Two inspectors each take their own random sample of 20 components and, independently, test \(\text{H}_0: p = 0.1\) against \(\text{H}_1: p > 0.1\) at the 5% level.(a) Find the critical region for each test.(b) Given that the factory's statement is true, find the probability that at least one of the inspectors rejects \(\text{H}_0\).(c) One inspector says, "Because we test at 5%, each test has a 5% chance of wrongly rejecting \(\text{H}_0\)." Comment on this statement.A* Challenge
Q14A coin is tossed \(n\) times to test \(\text{H}_0: p = 0.5\) against \(\text{H}_1: p > 0.5\), where \(p\) is the probability of a head. The critical region is \(X \geq n - 1\), where \(X\) is the number of heads.(a) Show that the actual significance level of the test is \(\dfrac{n + 1}{2^n}\).(b) Find the smallest value of \(n\) for which the actual significance level is at most 5%.(c) For this value of \(n\), show that \(X \geq n - 2\) could not be the critical region of a test at the 5% level.A* Challenge
Q15A person claims to be able to predict the suit of a card drawn at random from a standard pack. To test the claim, a card is drawn 40 times. Each time, the person predicts the suit, and the card is then replaced and the pack shuffled.(a) Explain why the card is replaced and the pack shuffled each time.(b) Using a 1% level of significance, find the smallest number of correct predictions that would support the person's claim, and the actual significance level.(c) The person makes only 4 correct predictions. A friend suggests that the person is deliberately choosing wrong answers. Test, at the 5% level, whether the proportion of correct predictions is less than would be expected from guessing.A* Challenge
(a) Let \(p\) be the proportion of light bulbs that are faulty. The claim gives \(\text{H}_0: p = 0.05\), and the shop owner believes it is higher: \(\text{H}_1: p > 0.05\). One direction is given, so the test is one-tailed
2
(b) Let \(p\) be the probability that the coin lands heads. \(\text{H}_0: p = 0.5\). "Biased" could mean too many heads or too few, so \(\text{H}_1: p \neq 0.5\), and the test is two-tailed
3
(c) Let \(p\) be the proportion of patients who recover with the new treatment. \(\text{H}_0: p = 0.7\). "Less effective" means fewer recover, so \(\text{H}_1: p < 0.7\), and the test is one-tailed
(a) \(\text{H}_0: p = 0.05\), \(\text{H}_1: p > 0.05\); one-tailed(b) \(\text{H}_0: p = 0.5\), \(\text{H}_1: p \neq 0.5\); two-tailed(c) \(\text{H}_0: p = 0.7\), \(\text{H}_1: p < 0.7\); one-tailed
Q2 — 0.0207 < 0.05: reject \(\text{H}_0\); evidence the coin is biased towards heads
1
Let \(p\) be the probability that the coin lands heads. \(\text{H}_0: p = 0.5\), \(\text{H}_1: p > 0.5\)
2
Let \(X\) be the number of heads. Assuming \(\text{H}_0\) is true, \(X \sim \text{B}(20, 0.5)\)
\(0.0207 < 0.05\), so the result is significant: reject \(\text{H}_0\). There is evidence, at the 5% level, that the coin is biased towards heads
Reject \(\text{H}_0\) (\(\text{P}(X \geq 15) = 0.0207\)): there is evidence that the coin is biased towards heads
Q3 — (a) \(X \leq 2\); (b) 0.0271; (c) Do not reject \(\text{H}_0\)
1
(a) Assuming \(\text{H}_0\) is true, \(X \sim \text{B}(15, 0.4)\). \(\text{H}_1\) is \(p < 0.4\), so the critical region is in the lower tail
2
\(\text{P}(X \leq 2) = 0.027114\ldots\), which is less than 0.05, and \(\text{P}(X \leq 3) = 0.090501\ldots\), which is more. So the critical region is \(X \leq 2\)
3
(b) The actual significance level is \(\text{P}(X \leq 2) = 0.027114\ldots\)
4
(c) 3 is not in the critical region, so do not reject \(\text{H}_0\). There is insufficient evidence, at the 5% level, that \(p < 0.4\)
(a) \(X \leq 2\)(b) 0.0271(c) Do not reject \(\text{H}_0\): insufficient evidence that \(p < 0.4\)
Q4 — (a) \(\text{H}_1: p \neq 0.5\); (b) 0.0386; (c) Do not reject \(\text{H}_0\)
1
(a) \(\text{H}_1\) is \(p \neq 0.5\), so too many heads or too few heads both count as evidence against \(\text{H}_0\). That is why the critical region has a tail at each end
(c) 9 is not in the critical region, so do not reject \(\text{H}_0\). There is insufficient evidence, at this significance level, that the coin is biased
(a) \(\text{H}_1\) is \(p \neq 0.5\), so both too many and too few heads lead to rejecting \(\text{H}_0\)(b) \(\dfrac{79}{2048} = 0.0386\)(c) Do not reject \(\text{H}_0\): insufficient evidence that the coin is biased
Q5 — 0.0905 > 0.05: do not reject \(\text{H}_0\); insufficient evidence
1
Let \(p\) be the proportion of households that recycle their food waste. \(\text{H}_0: p = 0.3\), \(\text{H}_1: p < 0.3\)
2
Let \(X\) be the number of households in the sample that recycle food waste. Assuming \(\text{H}_0\) is true, \(X \sim \text{B}(25, 0.3)\)
\(0.0905 > 0.05\), so the result is not significant: do not reject \(\text{H}_0\). There is insufficient evidence, at the 5% level, that fewer than 30% of households recycle their food waste
Do not reject \(\text{H}_0\) (\(\text{P}(X \leq 4) = 0.0905\)): there is insufficient evidence that fewer than 30% of households recycle food waste
Exam-Style (Q6–Q10)
Q6 — (a) \(X = 0\) or \(X \geq 10\); (b) 0.0211; (c) Do not reject \(\text{H}_0\)
1
(a) Let \(p\) be the proportion of seeds that produce white flowers. \(\text{H}_0: p = 0.2\), \(\text{H}_1: p \neq 0.2\). Assuming \(\text{H}_0\) is true, \(X \sim \text{B}(25, 0.2)\)
2
Lower tail: \(\text{P}(X = 0) = (0.8)^{25} = 0.003777\ldots\), which is less than 0.025, and \(\text{P}(X \leq 1) = 0.027389\ldots\), which is more. So the lower tail is \(X = 0\)
3
Upper tail: \(\text{P}(X \geq 9) = 1 - 0.953225\ldots = 0.046774\ldots\), which is more than 0.025
4
\(\text{P}(X \geq 10) = 1 - 0.982668\ldots = 0.017331\ldots\), which is less than 0.025. So the upper tail is \(X \geq 10\)
(c) 9 is not in the critical region, so do not reject \(\text{H}_0\). There is insufficient evidence, at the 5% level, that the proportion of seeds producing white flowers is not 20%
(a) \(X = 0\) or \(X \geq 10\)(b) 0.0211(c) Do not reject \(\text{H}_0\): insufficient evidence that the proportion is not 20%
Q7 — (a) 0.0132 > 0.01: do not reject \(\text{H}_0\); (b) Yes: 0.0132 < 0.05
1
(a) Let \(p\) be the proportion of visitors who buy a guidebook. \(\text{H}_0: p = 0.15\), \(\text{H}_1: p > 0.15\)
2
Let \(X\) be the number of visitors in the sample who buy a guidebook. Assuming \(\text{H}_0\) is true, \(X \sim \text{B}(50, 0.15)\)
\(0.0132 > 0.01\), so do not reject \(\text{H}_0\). There is insufficient evidence, at the 1% level, that the proportion of visitors who buy a guidebook has increased
5
(b) At the 5% level, \(0.0132 < 0.05\), so \(\text{H}_0\) would be rejected: there would be evidence that the proportion has increased
(a) Do not reject \(\text{H}_0\): insufficient evidence at the 1% level that the proportion has increased(b) Yes: \(0.0132 < 0.05\), so at the 5% level \(\text{H}_0\) would be rejected
Q8 — 0.0303 > 0.025: do not reject \(\text{H}_0\); insufficient evidence of a change
1
Let \(p\) be the proportion of customers who pay by direct debit. "Changed" gives no direction: \(\text{H}_0: p = 0.35\), \(\text{H}_1: p \neq 0.35\)
2
Assuming \(\text{H}_0\) is true, \(X \sim \text{B}(40, 0.35)\). You would expect about \(40 \times 0.35 = 14\) customers, and 8 is below this, so use the lower tail
3
\(\text{P}(X \leq 8) = 0.030254\ldots\)
4
The test is two-tailed, so compare with 0.025: \(0.0303 > 0.025\). Do not reject \(\text{H}_0\). There is insufficient evidence, at the 5% level, that the proportion paying by direct debit has changed
5
Note that \(0.0303 < 0.05\): a one-tailed test with \(\text{H}_1: p < 0.35\) would have rejected \(\text{H}_0\). Halving the significance level changes the conclusion here
Do not reject \(\text{H}_0\) (\(\text{P}(X \leq 8) = 0.0303 > 0.025\)): there is insufficient evidence that the proportion has changed
Q9 — (a) \(X \geq 13\); (b) 0.0216; (c) Do not reject \(\text{H}_0\)
1
(a) Let \(p\) be the proportion of patients who have side effects with the new version. \(\text{H}_0: p = 0.25\), \(\text{H}_1: p > 0.25\). Assuming \(\text{H}_0\) is true, \(X \sim \text{B}(30, 0.25)\)
2
\(\text{P}(X \geq 12) = 1 - \text{P}(X \leq 11) = 1 - 0.949341\ldots = 0.050658\ldots\), which is more than 0.05
3
\(\text{P}(X \geq 13) = 1 - \text{P}(X \leq 12) = 1 - 0.978406\ldots = 0.021593\ldots\), which is less than 0.05. So the critical region is \(X \geq 13\)
4
(b) The probability of incorrectly rejecting \(\text{H}_0\) is the actual significance level, \(\text{P}(X \geq 13) = 0.021593\ldots\)
5
(c) 12 is not in the critical region, so do not reject \(\text{H}_0\). There is insufficient evidence, at the 5% level, that the new version causes side effects in more than 25% of patients
(a) \(X \geq 13\)(b) 0.0216(c) Do not reject \(\text{H}_0\): insufficient evidence that more than 25% of patients have side effects
Q10 — (a) No direction is given; (b) 0.0210 < 0.05: reject \(\text{H}_0\); (c) Evidence, not proof
1
(a) The politician believes only that the figure is wrong, not whether it is too high or too low, so \(\text{H}_1: p \neq 0.6\)
2
(b) Let \(p\) be the proportion of adults who support the plan. \(\text{H}_0: p = 0.6\), \(\text{H}_1: p \neq 0.6\). Assuming \(\text{H}_0\) is true, \(X \sim \text{B}(20, 0.6)\)
3
You would expect about \(20 \times 0.6 = 12\) supporters, and 7 is below this, so use the lower tail: \(\text{P}(X \leq 7) = 0.021028\ldots\)
4
Compare with half of 10%: \(0.0210 < 0.05\). Reject \(\text{H}_0\). There is evidence, at the 10% level, that the proportion of adults who support the plan is not 60%
5
(c) A hypothesis test gives evidence, not proof. The sample might be unusual even if the newspaper's figure is correct: a test at the 10% level rejects a true \(\text{H}_0\) with probability up to 0.1
(a) The politician gives no direction for the change, so \(\text{H}_1: p \neq 0.6\)(b) Reject \(\text{H}_0\) (\(\text{P}(X \leq 7) = 0.0210 < 0.05\)): evidence that the proportion is not 60%(c) The test gives evidence, not proof: the conclusion could be wrong even though the test is carried out correctly
A* Challenge (Q11–Q15)
Q11 — (a) 0.0577 > 0.05: do not reject; (b) Reject at 10%; (c) Do not reject
1
(a) Let \(p\) be the probability of a head. \(\text{H}_0: p = 0.5\), \(\text{H}_1: p > 0.5\). Assuming \(\text{H}_0\) is true, \(X \sim \text{B}(20, 0.5)\)
\(0.0577 > 0.05\), so do not reject \(\text{H}_0\). There is insufficient evidence, at the 5% level, that the coin is biased towards heads
4
(b) \(0.0577 < 0.1\), so at the 10% level the result is significant: reject \(\text{H}_0\). There is evidence that the coin is biased towards heads
5
(c) With \(\text{H}_1: p \neq 0.5\), the 10% is split between the two tails. 14 is above the expected 10 heads, so compare the upper tail with 0.05: \(0.0577 > 0.05\). Do not reject \(\text{H}_0\): there is insufficient evidence that the coin is biased
6
The one-tailed test puts all 10% in the upper tail, but the two-tailed test puts only 5% there, so the same result needs to be more extreme to be significant
(a) Do not reject \(\text{H}_0\) (\(\text{P}(X \geq 14) = 0.0577 > 0.05\))(b) Reject \(\text{H}_0\): evidence that the coin is biased towards heads(c) Do not reject \(\text{H}_0\): the two-tailed test has only 5% in each tail
(a) \(\text{H}_0: p = 0.1\), \(\text{H}_1: p \neq 0.1\), where \(p\) is the proportion of defective items. Assuming \(\text{H}_0\) is true, \(X \sim \text{B}(10, 0.1)\), and each tail can have probability at most 0.025
2
The smallest possible lower tail is \(X = 0\), and \(\text{P}(X = 0) = (0.9)^{10} = 0.348678\ldots\), which is more than 0.025. So no value of \(X\) is in the lower tail
3
Upper tail: \(\text{P}(X \geq 3) = 1 - 0.929809\ldots = 0.070190\ldots\), which is more than 0.025, and \(\text{P}(X \geq 4) = 1 - 0.987204\ldots = 0.012795\ldots\), which is less. So the upper tail is \(X \geq 4\)
4
(b) Even a sample with no defective items would not be significant, so with 10 items the test can never give evidence that the proportion of defective items has decreased. Only an increase can be detected
5
(c) The lower tail is not empty when \(\text{P}(X = 0) = (0.9)^n \leq 0.025\). Taking logarithms, \(n \log 0.9 \leq \log 0.025\). Since \(\log 0.9\) is negative, dividing by it reverses the inequality: \(n \geq \dfrac{\log 0.025}{\log 0.9} = 35.01\ldots\)
6
Check either side: \((0.9)^{35} = 0.025031\ldots\), which is more than 0.025, and \((0.9)^{36} = 0.022528\ldots\), which is less. So \(n = 36\)
(a) No lower tail, because \(\text{P}(X = 0) = 0.3487\); upper tail \(X \geq 4\)(b) The test cannot give evidence that the proportion of defective items has decreased(c) \(n = 36\)
Q13 — (a) \(X \geq 5\); (b) 0.0845; (c) The actual level is 0.0432, not 5%
1
(a) Assuming \(\text{H}_0\) is true, \(X \sim \text{B}(20, 0.1)\). \(\text{P}(X \geq 4) = 1 - \text{P}(X \leq 3) = 1 - 0.867046\ldots = 0.132953\ldots\), which is more than 0.05
2
\(\text{P}(X \geq 5) = 1 - \text{P}(X \leq 4) = 1 - 0.956825\ldots = 0.043174\ldots\), which is less than 0.05. So the critical region is \(X \geq 5\)
3
(b) If \(\text{H}_0\) is true, each inspector does not reject \(\text{H}_0\) with probability \(\text{P}(X \leq 4) = 0.956825\ldots\)
4
The samples are independent, so \(\text{P}(\text{neither rejects}) = (0.956825\ldots)^2 = 0.915515\ldots\)
5
\(\text{P}(\text{at least one rejects}) = 1 - 0.915515\ldots = 0.084484\ldots\)
6
(c) \(X\) is discrete, so no critical region has probability exactly 0.05. The probability that each test wrongly rejects \(\text{H}_0\) is the actual significance level, 0.0432, which is less than 5%
(a) \(X \geq 5\)(b) 0.0845(c) The statement is not quite right: the probability of wrongly rejecting \(\text{H}_0\) is the actual significance level, 0.0432
\(\binom{n}{n - 1} = n\), so \(\text{P}(X \geq n - 1) = \dfrac{n + 1}{2^n}\)
4
(b) Going from \(n\) to \(n + 1\) multiplies the denominator by 2 but increases the numerator by only 1, so \(\dfrac{n + 1}{2^n}\) decreases as \(n\) increases
5
\(n = 7\): \(\dfrac{8}{128} = 0.0625\), which is more than 0.05. \(n = 8\): \(\dfrac{9}{256} = 0.035156\ldots\), which is less than 0.05. So \(n = 8\)
(a) Replacing and shuffling makes each draw independent of the others, and keeps the probability of guessing the suit correctly at \(\dfrac{1}{4}\) every time, so the number of correct predictions can be modelled by a binomial distribution
2
(b) Let \(p\) be the probability of a correct prediction. \(\text{H}_0: p = 0.25\), \(\text{H}_1: p > 0.25\). Assuming \(\text{H}_0\) is true, \(X \sim \text{B}(40, 0.25)\)
3
\(\text{P}(X \geq 17) = 1 - \text{P}(X \leq 16) = 1 - 0.988438\ldots = 0.011561\ldots\), which is more than 0.01
4
\(\text{P}(X \geq 18) = 1 - \text{P}(X \leq 17) = 1 - 0.995348\ldots = 0.004651\ldots\), which is less than 0.01. So at least 18 correct predictions are needed, and the actual significance level is 0.004651…
5
(c) Now \(\text{H}_0: p = 0.25\), \(\text{H}_1: p < 0.25\), with \(X \sim \text{B}(40, 0.25)\) assuming \(\text{H}_0\) is true. Lower tail: \(\text{P}(X \leq 4) = 0.016042\ldots\)
6
\(0.0160 < 0.05\), so reject \(\text{H}_0\). There is evidence, at the 5% level, that the person's proportion of correct predictions is less than 0.25, which supports the friend's suggestion
(a) So that the draws are independent, with the same probability \(\dfrac{1}{4}\) of a correct guess each time(b) 18 correct predictions; actual significance level 0.0047(c) Reject \(\text{H}_0\) (\(\text{P}(X \leq 4) = 0.0160 < 0.05\)): evidence that the proportion of correct predictions is less than 0.25
Summary
\(\text{H}_0\) gives a single value of the population parameter \(p\); \(\text{H}_1\) is \(p > \ldots\) or \(p < \ldots\) (one-tailed) or \(p \neq \ldots\) (two-tailed). Define \(p\) in context.
Assuming \(\text{H}_0\) is true, the test statistic \(X\), the number of successes in the sample, has distribution \(\text{B}(n, p)\).
Find the probability of the observed value or anything more extreme, and reject \(\text{H}_0\) if it is less than or equal to the significance level.
The critical region holds every value that leads to rejecting \(\text{H}_0\). Its probability is the actual significance level, which is the probability of incorrectly rejecting \(\text{H}_0\).
In a two-tailed test, each tail has at most half the significance level. Always conclude in context: "there is (insufficient) evidence that…".
If setting up hypotheses, finding critical regions or writing conclusions in context are still causing problems, Alamin's diagnostic approach identifies exactly which skills are missing and builds a targeted plan to address them — with AI-powered practice between sessions.