Scribe logoScribeNotes & practice
Sign inStart free
  1. Notes
  2. ›AP Statistics
  3. ›3 Inference for Categorical Data: Proportions
  4. ›3.5 Setting up a test for a proportion

AP Statistics · Unit 3 · Inference for Categorical Data: Proportions

3.5 Setting up a test for a proportion

State hypotheses for a population proportion, check the conditions for a test, and choose the correct procedure for a proportion scenario.

  • 5 min read
  • 5 parts
  • 10 questions
  • Draft
Start a Quick 5 on this topic→Jump to the 10 questions
1/5A significance test checks a claim using sample data
  1. 1A significance test checks a claim using sample data
  2. 2Hypotheses describe the population proportion, not the sample
  3. 3The question’s wording decides the form of the alternative
  4. 4Three conditions must hold before a normal model is used
  5. 5The data type and question decide which test to use
  6. Check your recall
  7. Practice: 10 questions
1

Part 1 of 5: A significance test checks a claim using sample data

Earlier, an interval was used to judge a claim. A test judges a claim more directly, with a clear rule for when the evidence is strong. Suppose a company says that 30% of its customers renew each year. You sample customers and see a lower renewal rate in the sample.

A significance test asks whether that gap is bigger than random sampling alone would usually produce. It begins by setting up two competing statements about the population. No calculation happens until the setup is correct. A weak setup cannot be repaired later by careful arithmetic.

Key term

significance test

A significance test is a procedure that uses sample data to judge a claim about a population proportion by measuring evidence against it.

Part 1 done

2

Part 2 of 5: Hypotheses describe the population proportion, not the sample

Now state the two competing claims. The null hypothesis, written H0H_0H0​, says the population proportion equals one specific value. The alternative hypothesis, written HaH_aHa​, says the proportion differs from that value.

Let ppp be the population proportion and p0p_0p0​ the claimed value, called the null value. The sample proportion, p^\hat{p}p^​, is the fraction of the sample with the trait. The null hypothesis is then H0:p=p0H_0: p = p_0H0​:p=p0​.

Common mistake

Write hypotheses with ppp, never with the sample proportion p^\hat{p}p^​. The sample result is evidence, not part of the claim.

Key terms · tap one for its definition

null hypothesis

The null hypothesis is the claim tested by a significance test, stating that the population proportion equals a specific value, p0p_0p0​.

alternative hypothesis

The alternative hypothesis is the claim that the population proportion differs from the null value, which the data are used to support.

Part 2 done

Quick check

One question on the part you just read

Easy

A manager claims that 70% of customers are satisfied and wants to test whether the proportion differs from 0.70. Identify the correct hypotheses.
Show answer and explanation

Answer: A · Worked explanation

The null hypothesis holds the claimed value with an equals sign, and a difference in either direction gives a two-sided alternative. The version using p^\hat{p}p^​ is wrong because hypotheses describe the population proportion, not the sample proportion. Swapping the null and alternative statements is also wrong, because the null always holds the equals sign.
3

Part 3 of 5: The question’s wording decides the form of the alternative

Next, decide the direction. Read the question for the researcher’s suspicion before looking at any data. Two forms are possible, and the choice affects how the evidence is judged later.

  • A one-sided alternative, Ha:p<p0H_a: p < p_0Ha​:p<p0​ or Ha:p>p0H_a: p > p_0Ha​:p>p0​, names a direction such as lower or greater.
  • A two-sided alternative, Ha:p≠p0H_a: p \neq p_0Ha​:p=p0​, allows a difference in either direction.

Worked example

Writing hypotheses

A coffee shop claims that 60% of its customers order a drink with milk. The manager believes the true proportion is lower. Write the hypotheses.
  1. 1
    Define the parameter: ppp is the proportion of all customers who order a drink with milk.
  2. 2
    The null value comes from the claim, so H0:p=0.60H_0: p = 0.60H0​:p=0.60.
  3. 3
    The manager suspects a lower proportion, so the alternative is one-sided: Ha:p<0.60H_a: p < 0.60Ha​:p<0.60.

Answer: H0:p=0.60H_0: p = 0.60H0​:p=0.60 and Ha:p<0.60H_a: p < 0.60Ha​:p<0.60, where ppp is the proportion of all customers who order a drink with milk.

The null value comes from the claim, and the direction comes from the suspicion. The sample result plays no part in either choice. Choosing the direction after seeing the data would bias the test.

Key terms · tap one for its definition

one-sided alternative

A one-sided alternative is an alternative hypothesis that states the population proportion differs from the null value in one specified direction.

two-sided alternative

A two-sided alternative is an alternative hypothesis that states the population proportion differs from the null value in either direction.

Part 3 done

4

Part 4 of 5: Three conditions must hold before a normal model is used

Once the hypotheses are set, check that the test is valid. These checks let you model the possible sample proportions with a normal curve. If a check fails, the test results may not be reliable. Let nnn be the sample size.

  • The sample is random, or the data come from a randomized experiment.
  • The 10% condition holds: the sample is no more than 10% of the population, so observations are nearly independent.
  • The large counts condition holds: np0≥10np_0 \geq 10np0​≥10 and n(1−p0)≥10n(1-p_0) \geq 10n(1−p0​)≥10.

Worked example

Checking conditions

A random sample of 150 customers is taken from 5,000 customers. Check the conditions for testing H0:p=0.30H_0: p = 0.30H0​:p=0.30.
  1. 1
    Random: the sample is random, as given.
  2. 2
    10%: 10% of 5,000 is 500, and 150 is less than 500.
  3. 3
    Large counts: 150(0.30)=45150(0.30) = 45150(0.30)=45 and 150(0.70)=105150(0.70) = 105150(0.70)=105, and both are at least 10.

Answer: All three conditions are met, so a normal model is appropriate.

Exam tip

In a test, check large counts with the claimed value p0p_0p0​. Do not use the sample proportion p^\hat{p}p^​ here.

Key terms · tap one for its definition

10% condition

The 10% condition requires that a sample taken without replacement be no more than 10% of the population, so observations are nearly independent.

large counts condition

The large counts condition requires that the expected successes np0np_0np0​ and failures n(1−p0)n(1-p_0)n(1−p0​) are both at least 10.

Part 4 done

Quick check

One question on the part you just read

Easy

Which wording signals a two-sided alternative hypothesis?
Show answer and explanation

Answer: C · Worked explanation

Asking whether a proportion has changed allows a difference in either direction, which is a two-sided alternative. Increased, lower, and exceeds each name one direction, so each leads to a one-sided alternative. Only the word changed leaves both directions open, and that is the feature to look for in a question.
5

Part 5 of 5: The data type and question decide which test to use

Finally, choose the procedure. Ask what kind of data you have and what the question needs. The answers narrow the choices quickly.

A one-sample z-test for a proportion tests a claimed proportion. It uses one random sample and a categorical variable with two outcomes. Its name refers to the z statistic. This statistic counts how many standard deviations the sample proportion lies from the claimed value.

Matching a scenario to a procedure

ScenarioProcedure
One sample, two outcomes, a claimed proportionOne-sample z-test for a proportion
Estimate a proportion with no claim to testConfidence interval for a proportion
Quantitative data, such as heightsA test for a mean, not a proportion
Scenario
One sample, two outcomes, a claimed proportion
Procedure
One-sample z-test for a proportion
Scenario
Estimate a proportion with no claim to test
Procedure
Confidence interval for a proportion
Scenario
Quantitative data, such as heights
Procedure
A test for a mean, not a proportion

Exam tip

Look for three clues: one categorical variable, one sample, and a claimed value. Together they point to a one-sample z-test for a proportion.

Key term

one-sample z-test for a proportion

A one-sample z-test for a proportion is a significance test of a claimed population proportion that uses a z statistic from one random sample.

Note read

Check your recall[12 marks]

5 written questions. Answer each one in your head or on paper, then open the mark scheme.

  1. 1.Identify[3 marks]

    A bakery claims that 25% of its customers buy a pastry. The owner believes the true proportion is higher. Identify the null and alternative hypotheses, defining the parameter.

    Show mark schemeHide mark scheme

    Model answer

    Let ppp be the proportion of all customers who buy a pastry. H0:p=0.25H_0: p = 0.25H0​:p=0.25 and Ha:p>0.25H_a: p > 0.25Ha​:p>0.25.

    • 1 mark for defining ppp as the proportion of all customers who buy a pastry
    • 1 mark for writing H0:p=0.25H_0: p = 0.25H0​:p=0.25
    • 1 mark for writing Ha:p>0.25H_a: p > 0.25Ha​:p>0.25
  2. 2.Explain[2 marks]

    Explain why hypotheses are written about the population proportion rather than the sample proportion.

    Show mark schemeHide mark scheme

    Model answer

    A test judges a claim about the whole population. The sample proportion is only the evidence collected, so it is not part of the claim.

    • 1 mark for stating that the claim concerns the population
    • 1 mark for stating that the sample proportion is the evidence, not part of the claim
  3. 3.Determine[2 marks]

    Determine whether the large counts condition is met for a random sample of 60 when H0:p=0.10H_0: p = 0.10H0​:p=0.10.

    Show mark schemeHide mark scheme

    Model answer

    60(0.10)=660(0.10) = 660(0.10)=6 and 60(0.90)=5460(0.90) = 5460(0.90)=54. The expected number of successes, 6, is less than 10, so the condition is not met.

    • 1 mark for calculating both expected counts, 6 and 54
    • 1 mark for concluding that the condition is not met because 6 is less than 10
  4. 4.Justify[2 marks]

    A random sample of 200 voters is asked a yes or no question to test a claim that 55% of all voters in a city agree. Justify the choice of a one-sample z-test for a proportion.

    Show mark schemeHide mark scheme

    Model answer

    There is one random sample and one categorical variable with two outcomes. The question compares the proportion with a claimed value.

    • 1 mark for identifying one random sample with a two-outcome categorical variable
    • 1 mark for linking the test to comparing a proportion with a claimed value
  5. 5.Describe[3 marks]

    Describe the three conditions to check before using a one-sample z-test for a proportion.

    Show mark schemeHide mark scheme

    Model answer

    The sample must be random. The sample must be no more than 10% of the population. Both np0np_0np0​ and n(1−p0)n(1-p_0)n(1−p0​) must be at least 10.

    • 1 mark for stating that the sample is random
    • 1 mark for stating the 10% condition
    • 1 mark for stating that np0np_0np0​ and n(1−p0)n(1-p_0)n(1−p0​) are both at least 10

Practice this topic

10 quick-fire multiple-choice questions, easy to hard. Tap an answer to see if you're right and the worked explanation. (The written questions with mark schemes are under “Check your recall” above.)

0 right0 of 10 answered

Includes the 2 quick checks in the note above.

3.

Easy·Identify[1 mark]

Identify the condition that requires the sample to be no more than 10% of the population.
Show answer and explanation

Answer: B · Worked explanation

The 10% condition limits the sample size relative to the population so that observations are nearly independent when sampling without replacement. The large counts condition is about expected successes and failures, and the random condition is about how the sample was chosen.
4.

Medium[1 mark]

A random sample of n=120n = 120n=120 is used to test H0:p=0.15H_0: p = 0.15H0​:p=0.15. Calculate the expected numbers of successes and failures used in the large counts check.
Show answer and explanation

Answer: D · Worked explanation

Expected successes are 120(0.15)=18120(0.15) = 18120(0.15)=18, and expected failures are 120(0.85)=102120(0.85) = 102120(0.85)=102. Both are at least 10. The sample size, 120, is not a count of successes, and failures must use 1−p0=0.851 - p_0 = 0.851−p0​=0.85, not 0.15 again. The two expected counts must also add up to 120.
5.

Medium[1 mark]

A random sample of 40 is taken, without replacement, from a population of 300 to test H0:p=0.50H_0: p = 0.50H0​:p=0.50. Determine which condition is not met.
Show answer and explanation

Answer: A · Worked explanation

Ten percent of 300 is 30, and the sample of 40 exceeds that, so the 10% condition fails. The large counts condition holds because 40(0.50)=2040(0.50) = 2040(0.50)=20 and the failures also number 20, and both are at least 10. Sample size alone does not decide the random condition.
6.

Medium[1 mark]

A manager claims that 8% of items are defective and tests this claim with a random sample of 90 items. Determine whether the large counts condition is met.
Show answer and explanation

Answer: C · Worked explanation

Both expected counts must be at least 10. The expected successes are 90(0.08)=7.290(0.08) = 7.290(0.08)=7.2, which is below 10, so the condition fails. Checking only the failures, 82.8, misses this problem. A sample above 30 does not replace the check, because the large counts condition is the one that matters for proportions.
7.

Medium·Identify[1 mark]

Identify the scenario that calls for a one-sample z-test for a proportion.
Show answer and explanation

Answer: B · Worked explanation

The test applies to one categorical variable from one sample and a claimed proportion. A mean commute time involves quantitative data, and two classes involve two samples. Estimating with no claim calls for an interval rather than a test, since there is no claimed value to compare with.
8.

Hard[2 marks]

A student writes H0:p^=0.40H_0: \hat{p} = 0.40H0​:p^​=0.40 and Ha:p^>0.40H_a: \hat{p} > 0.40Ha​:p^​>0.40 after finding that 46% of a sample have a trait. Identify the main error.
Show answer and explanation

Answer: D · Worked explanation

Hypotheses are statements about the population proportion ppp, so they cannot contain p^\hat{p}p^​. The null value comes from the claim, which is 0.40, not from the data. The null hypothesis always uses equality, and the alternative’s form depends on the question.
9.

Hard[2 marks]

A random sample of 50 is used to test H0:p=0.10H_0: p = 0.10H0​:p=0.10, and the sample proportion is 0.30. Determine whether the large counts condition is met.
Show answer and explanation

Answer: A · Worked explanation

A test assumes the null is true, so the counts use p0=0.10p_0 = 0.10p0​=0.10. That gives 50(0.10)=550(0.10) = 550(0.10)=5, which is below 10, so the condition fails. Using the sample proportion gives 15 and 35, which wrongly suggests the condition passes.
10.

Hard[2 marks]

A researcher sees a sample proportion of 0.62 and then chooses Ha:p>0.50H_a: p > 0.50Ha​:p>0.50 because the sample is above 0.50. What is the best criticism?
Show answer and explanation

Answer: C · Worked explanation

Choosing the direction after looking at the data builds the result into the test and makes the evidence look stronger than it is. The null value still comes from the claim, 0.50. A one-sided alternative is allowed when the question names a direction in advance.

Keep going

Next topic3.6 p-valuesPrevious topic3.4 Claims from a proportion interval

More in Unit 3 · Inference for Categorical Data: Proportions

  1. 3.1EstimatorsNotes · 10 questions
  2. 3.2Sample proportionsNotes · 9 questions
  3. 3.3Confidence interval for a proportionNotes · 10 questions
  4. 3.4Claims from a proportion intervalNotes · 10 questions
  5. 3.5Setting up a test for a proportionNotes · 10 questions
  6. 3.6p-valuesNotes · 10 questions
  7. 3.7Carrying out a test for a proportionNotes · 10 questions
  8. 3.8Errors in testsNotes · 10 questions
  9. 3.9Differences in sample proportionsNotes · 10 questions

All AP Statistics topics →

Draft: written by Scribe and checked automatically, not yet by a subject reviewer.

Practice questions are original, written by Scribe and checked automatically for correctness.

Topic codes follow the official numbering. Not endorsed by College Board or any exam board.

On this page

0/5 read

  1. 1A significance test checks a claim using sample data
  2. 2Hypotheses describe the population proportion, not the sample
  3. 3The question’s wording decides the form of the alternative
  4. 4Three conditions must hold before a normal model is used
  5. 5The data type and question decide which test to use
  6. Check your recall
  7. Practice: 10 questions
All coursesScribeHow it worksPricingPrivacyTerms

Notes and questions are written by Scribe in our own words. Topic codes follow the official course numbering so you can find your place. Not produced or endorsed by the IB, College Board, AQA, OCR, Pearson Edexcel or any exam board.

© 2026 Scribe