Scribe logoScribeNotes & practice
Sign inStart free
  1. Notes
  2. ›AP Statistics
  3. ›3 Inference for Categorical Data: Proportions
  4. ›3.9 Differences in sample proportions

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

3.9 Differences in sample proportions

Calculate the mean and standard deviation of the difference of two sample proportions, describe its sampling distribution, and check the conditions for a normal approximation.

  • 4 min read
  • 5 parts
  • 10 questions
  • Draft
Start a Quick 5 on this topic→Jump to the 10 questions
1/5Comparing two groups starts with a difference of sample proportions
  1. 1Comparing two groups starts with a difference of sample proportions
  2. 2The distribution is centered at the true difference in proportions
  3. 3The standard deviation combines the variability of both samples
  4. 4Large samples make the distribution approximately normal
  5. 5Four conditions must hold before using the normal model
  6. Check your recall
  7. Practice: 10 questions
1

Part 1 of 5: Comparing two groups starts with a difference of sample proportions

Two cities each take a random sample of residents. Each sample gives the proportion who support a new bike lane. We often want to compare the cities, not describe one alone.

Subtract the second sample proportion from the first. The result is the difference in sample proportions, p^1−p^2\hat{p}_1-\hat{p}_2p^​1​−p^​2​. Here p^1\hat{p}_1p^​1​ and p^2\hat{p}_2p^​2​ are the proportions of successes in samples of sizes n1n_1n1​ and n2n_2n2​.

This statistic estimates the true difference p1−p2p_1-p_2p1​−p2​ between the population proportions. A new pair of samples gives a new difference, so the statistic varies. The pattern of its values over all possible pairs of samples is its sampling distribution.

Exam tip

Pick an order and keep it. If group 1 comes first in p1−p2p_1-p_2p1​−p2​, it must come first everywhere else.

Key term

difference in sample proportions

The difference in sample proportions is p^1−p^2\hat{p}_1-\hat{p}_2p^​1​−p^​2​, computed from independent random samples of sizes n1n_1n1​ and n2n_2n2​ taken from two populations.

Part 1 done

2

Part 2 of 5: The distribution is centered at the true difference in proportions

Now consider where these differences cluster. Each sample proportion is centered on its own population proportion. So the difference is centered on p1−p2p_1-p_2p1​−p2​.

In symbols, the mean of the sampling distribution is μp^1−p^2=p1−p2\mu_{\hat{p}_1-\hat{p}_2}=p_1-p_2μp^​1​−p^​2​​=p1​−p2​. On average, the statistic neither overshoots nor undershoots the true difference. In plain terms, the statistic has no built-in lean toward larger or smaller values.

Suppose p1=0.60p_1=0.60p1​=0.60 and p2=0.50p_2=0.50p2​=0.50. The center is 0.60−0.50=0.100.60-0.50=0.100.60−0.50=0.10. Individual differences land above or below 0.10 by chance. Another random pair of samples might give 0.07 or 0.13 instead.

Part 2 done

Quick check

One question on the part you just read

Easy

Identify the mean of the sampling distribution of p^1−p^2\hat{p}_1-\hat{p}_2p^​1​−p^​2​.
Show answer and explanation

Answer: B · Worked explanation

Each sample proportion is centered on its own population proportion, so their difference is centered on p1−p2p_1-p_2p1​−p2​. The sum p1+p2p_1+p_2p1​+p2​ would apply to adding the proportions, not subtracting them. A mean of 0 holds only when the two population proportions are equal.
3

Part 3 of 5: The standard deviation combines the variability of both samples

Next comes the spread. Each sample proportion has standard deviation p(1−p)n\sqrt{\frac{p(1-p)}{n}}np(1−p)​​. The variance is the square of the standard deviation, and variances add when the two samples are independent samples. That means knowing one sample tells us nothing about the other.

Add the two variances, then take the square root. This gives σp^1−p^2=p1(1−p1)n1+p2(1−p2)n2\sigma_{\hat{p}_1-\hat{p}_2}=\sqrt{\frac{p_1(1-p_1)}{n_1}+\frac{p_2(1-p_2)}{n_2}}σp^​1​−p^​2​​=n1​p1​(1−p1​)​+n2​p2​(1−p2​)​​. Each term under the root belongs to one sample. A larger sample makes its term smaller, which shrinks the overall spread.

Common mistake

Do not subtract the standard deviations. The two sources of variability combine, so the spread of the difference exceeds the spread of either sample proportion alone.

Worked example

Mean and standard deviation of a difference

Two independent random samples of size n1=n2=100n_1=n_2=100n1​=n2​=100 come from populations with p1=0.60p_1=0.60p1​=0.60 and p2=0.50p_2=0.50p2​=0.50. Calculate the mean and standard deviation of p^1−p^2\hat{p}_1-\hat{p}_2p^​1​−p^​2​.
  1. 1
    Mean: μ=p1−p2=0.60−0.50=0.10\mu=p_1-p_2=0.60-0.50=0.10μ=p1​−p2​=0.60−0.50=0.10.
  2. 2
    Standard deviation is the square root of the sum of the two variances: σ=0.60(0.40)100+0.50(0.50)100\sigma=\sqrt{\frac{0.60(0.40)}{100}+\frac{0.50(0.50)}{100}}σ=1000.60(0.40)​+1000.50(0.50)​​.
  3. 3
    Inside the root: 0.0024+0.0025=0.00490.0024+0.0025=0.00490.0024+0.0025=0.0049.
  4. 4
    Take the root: 0.0049=0.07\sqrt{0.0049}=0.070.0049​=0.07.

Answer: The mean is 0.10 and the standard deviation is 0.07.

Key term

independent samples

Independent samples are samples in which the individuals chosen for one group have no influence on, or connection to, the individuals chosen for the other.

Part 3 done

4

Part 4 of 5: Large samples make the distribution approximately normal

With the center and spread known, consider the shape. When both samples are large enough, the sampling distribution is approximately normal. This normal approximation lets us use zzz-scores and normal probabilities for the difference.

Figure 1. The distribution of the difference is bell-shaped, centered at 0.10, with a standard deviation of 0.07.

In the figure, the curve is centered at 0.10 with standard deviation 0.07. By the empirical rule, about 95% of differences fall within two standard deviations. That is roughly from -0.04 to 0.24.

Without large samples, the distribution can be skewed, and normal probabilities may mislead.

Key term

normal approximation

A normal approximation is the use of a normal distribution in place of a statistic’s exact sampling distribution when sample sizes are large enough.

Part 4 done

Quick check

One question on the part you just read

Easy

Identify the shape of the sampling distribution of p^1−p^2\hat{p}_1-\hat{p}_2p^​1​−p^​2​ when the conditions are met.
Show answer and explanation

Answer: C · Worked explanation

When the random, 10%, and Large Counts conditions hold, the sampling distribution is approximately normal. Skewness is a concern only when the counts are small, which is exactly what the Large Counts condition guards against. A uniform or bimodal shape does not arise from these samples.
5

Part 5 of 5: Four conditions must hold before using the normal model

Before trusting that bell shape, check the conditions in order.

  • Random: both samples are random, or treatments were randomly assigned.
  • 10% condition: each sample is less than 10% of its population, so observations are nearly independent.
  • Large Counts condition: n1p1n_1p_1n1​p1​, n1(1−p1)n_1(1-p_1)n1​(1−p1​), n2p2n_2p_2n2​p2​, and n2(1−p2)n_2(1-p_2)n2​(1−p2​) are all at least 10.
  • Independent: individuals in one sample are unrelated to those in the other.

When p1p_1p1​ and p2p_2p2​ are unknown, use the sample proportions to compute the four counts.

Exam tip

Show the four counts with numbers. Write each product and compare it with 10.

Worked example

Checking the conditions

Both samples have 100 individuals, drawn at random and independently from populations of at least 2,000 people. The proportions are p1=0.60p_1=0.60p1​=0.60 and p2=0.50p_2=0.50p2​=0.50. Check the conditions.
  1. 1
    Random and independent: the samples are random, and each comes from a separate population.
  2. 2
    10% condition: 100 is less than 10% of 2,000, which is 200.
  3. 3
    Large Counts: 100(0.60)=60100(0.60)=60100(0.60)=60, 100(0.40)=40100(0.40)=40100(0.40)=40, 100(0.50)=50100(0.50)=50100(0.50)=50, and 100(0.50)=50100(0.50)=50100(0.50)=50.
  4. 4
    All four counts are at least 10.

Answer: All conditions are met, so the normal model is reasonable.

Key terms · tap one for its definition

10% condition

The 10% condition is met when each sample is less than 10% of its population, so sampled individuals can be treated as approximately independent.

Large Counts condition

The Large Counts condition is met when each group has at least 10 expected successes and at least 10 expected failures.

Note read

Check your recall[12 marks]

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

  1. 1.Identify[1 mark]

    Identify the mean of the sampling distribution of p^1−p^2\hat{p}_1-\hat{p}_2p^​1​−p^​2​ in terms of p1p_1p1​ and p2p_2p2​.

    Show mark schemeHide mark scheme

    Model answer

    The mean is p1−p2p_1-p_2p1​−p2​.

    • 1 mark for identifying the mean as p1−p2p_1-p_2p1​−p2​
  2. 2.Calculate[3 marks]

    Two independent random samples have n1=80n_1=80n1​=80, p1=0.30p_1=0.30p1​=0.30, n2=120n_2=120n2​=120, and p2=0.20p_2=0.20p2​=0.20. Calculate the mean and standard deviation of p^1−p^2\hat{p}_1-\hat{p}_2p^​1​−p^​2​.

    Show mark schemeHide mark scheme

    Model answer

    The mean is 0.30−0.20=0.100.30-0.20=0.100.30−0.20=0.10. The standard deviation is 0.30(0.70)80+0.20(0.80)120=0.002625+0.001333≈0.063\sqrt{\frac{0.30(0.70)}{80}+\frac{0.20(0.80)}{120}}=\sqrt{0.002625+0.001333}\approx0.063800.30(0.70)​+1200.20(0.80)​​=0.002625+0.001333​≈0.063.

    • 1 mark for calculating the mean as 0.10
    • 1 mark for substituting both variance terms correctly under one square root
    • 1 mark for obtaining a standard deviation of about 0.063
  3. 3.Describe[2 marks]

    Describe the shape and center of the sampling distribution of p^1−p^2\hat{p}_1-\hat{p}_2p^​1​−p^​2​ when all conditions are met.

    Show mark schemeHide mark scheme

    Model answer

    The distribution is approximately normal and is centered at p1−p2p_1-p_2p1​−p2​.

    • 1 mark for stating that the shape is approximately normal
    • 1 mark for stating that the center is p1−p2p_1-p_2p1​−p2​
  4. 4.Explain[2 marks]

    Explain why the standard deviation of p^1−p^2\hat{p}_1-\hat{p}_2p^​1​−p^​2​ is found by adding variances rather than subtracting standard deviations.

    Show mark schemeHide mark scheme

    Model answer

    For independent samples, variances add because variability from both samples affects the difference. Subtracting would wrongly suggest that the difference varies less.

    • 1 mark for stating that variances of independent samples add
    • 1 mark for explaining that both samples contribute variability, so the spread is not reduced by subtraction
  5. 5.Determine[2 marks]

    Determine whether the Large Counts condition is met when n1=50n_1=50n1​=50, p1=0.40p_1=0.40p1​=0.40, n2=60n_2=60n2​=60, and p2=0.15p_2=0.15p2​=0.15.

    Show mark schemeHide mark scheme

    Model answer

    The four counts are 50(0.40)=2050(0.40)=2050(0.40)=20, 50(0.60)=3050(0.60)=3050(0.60)=30, 60(0.15)=960(0.15)=960(0.15)=9, and 60(0.85)=5160(0.85)=5160(0.85)=51. Since 9 is below 10, the condition is not met.

    • 1 mark for computing the four counts as 20, 30, 9, and 51
    • 1 mark for concluding that the condition is not met because 9 is below 10
  6. 6.Justify[2 marks]

    A teacher asks the same 40 students about two different topics, then treats the two sets of answers as two independent samples. Justify why this is not appropriate.

    Show mark schemeHide mark scheme

    Model answer

    The same individuals appear in both samples, so the samples are not independent. The formula that adds variances therefore does not apply.

    • 1 mark for stating that the same students are in both samples
    • 1 mark for concluding that the samples are not independent, so the standard deviation formula does not apply

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 standard deviation of p^1−p^2\hat{p}_1-\hat{p}_2p^​1​−p^​2​ for independent samples.
Show answer and explanation

Answer: D · Worked explanation

Variances of independent samples add, and the standard deviation is the square root of that sum. The version without a root is the variance, not the standard deviation. Subtracting standard deviations or subtracting variances is wrong because both samples add variability.
4.

Medium·Calculate[2 marks]

Two independent random samples have n1=100n_1=100n1​=100, p1=0.50p_1=0.50p1​=0.50, n2=150n_2=150n2​=150, and p2=0.40p_2=0.40p2​=0.40. Calculate the standard deviation of p^1−p^2\hat{p}_1-\hat{p}_2p^​1​−p^​2​.
Show answer and explanation

Answer: C · Worked explanation

The variances are 0.50(0.50)100=0.0025\frac{0.50(0.50)}{100}=0.00251000.50(0.50)​=0.0025 and 0.40(0.60)150=0.0016\frac{0.40(0.60)}{150}=0.00161500.40(0.60)​=0.0016. Their sum is 0.0041, and the square root is about 0.064. The value 0.0041 forgets the root, and 0.030 comes from subtracting the variances, 0.0025−0.0016\sqrt{0.0025-0.0016}0.0025−0.0016​. The value 0.10 is just the difference of the two population proportions, which is the mean, not the spread.
5.

Medium·Identify[2 marks]

Identify the pair of samples that satisfies the Large Counts condition.
Show answer and explanation

Answer: D · Worked explanation

For the correct pair the counts are 30, 30, 20, and 60, all at least 10. The first pair fails because 50(0.15)=7.550(0.15)=7.550(0.15)=7.5. The second fails because 40(0.10)=440(0.10)=440(0.10)=4 failures. The third fails because 200(0.04)=8200(0.04)=8200(0.04)=8 successes. Every one of the four counts in each pair must reach 10, so a single small count is enough to break the condition.
6.

Medium·Identify[2 marks]

A random sample of 150 residents comes from a town of 1,200, and a random sample of 80 residents comes from a city of 5,000. Identify which statement about the 10% condition is correct.
Show answer and explanation

Answer: D · Worked explanation

Ten percent of 1,200 is 120, and 150 exceeds 120, so the first sample fails. Ten percent of 5,000 is 500, and 80 is below it, so the second sample passes. A sample size above 30 does not meet this condition, because it compares sample size with population size.
7.

Medium·Identify[2 marks]

Identify the design that produces independent samples.
Show answer and explanation

Answer: B · Worked explanation

Samples are independent when the individuals in one group have no connection to those in the other, as with separate random samples from different schools. Measuring the same students twice links the groups. Siblings and twins are related, so choosing one group informs the other.
8.

Hard·Predict[2 marks]

Both sample sizes are multiplied by 4, and the population proportions stay the same. Predict what happens to the standard deviation of p^1−p^2\hat{p}_1-\hat{p}_2p^​1​−p^​2​.
Show answer and explanation

Answer: C · Worked explanation

Each variance term is divided by 4, so the sum is divided by 4. The standard deviation is the square root, so it is divided by 4=2\sqrt{4}=24​=2. Dividing by 4 is the tempting error, because it treats the variance change as a standard deviation change.
9.

Hard[2 marks]

Which statement about the standard deviation of p^1−p^2\hat{p}_1-\hat{p}_2p^​1​−p^​2​ for independent samples is true?
Show answer and explanation

Answer: D · Worked explanation

The variances add, so the standard deviation of the difference is σ12+σ22\sqrt{\sigma_1^2+\sigma_2^2}σ12​+σ22​​, which exceeds both σ1\sigma_1σ1​ and σ2\sigma_2σ2​. Subtracting the standard deviations ignores that both samples add variability. Adding the standard deviations overstates the spread, because the square root of a sum is smaller than the sum of roots.
10.

Hard·Determine[2 marks]

Independent random samples have n1=50n_1=50n1​=50, p1=0.84p_1=0.84p1​=0.84, n2=50n_2=50n2​=50, and p2=0.60p_2=0.60p2​=0.60, and both samples are less than 10% of their populations. Determine whether the normal model is appropriate.
Show answer and explanation

Answer: D · Worked explanation

The first group has 50(0.84)=4250(0.84)=4250(0.84)=42 successes and 50(1−0.84)=850(1-0.84)=850(1−0.84)=8 failures. Eight is below 10, so the Large Counts condition fails. The second group has 30 successes and 20 failures, which are fine. Combining counts or citing a size of 30 does not satisfy the condition.

Keep going

Next topic3.10 Interval for a difference of proportionsPrevious topic3.8 Errors in tests

More in Unit 3 · Inference for Categorical Data: Proportions

  1. 3.5Setting up a test for a proportionNotes · 10 questions
  2. 3.6p-valuesNotes · 10 questions
  3. 3.7Carrying out a test for a proportionNotes · 10 questions
  4. 3.8Errors in testsNotes · 10 questions
  5. 3.9Differences in sample proportionsNotes · 10 questions
  6. 3.10Interval for a difference of proportionsNotes · 10 questions
  7. 3.11Claims from a difference intervalNotes · 10 questions
  8. 3.12Setting up a two-proportion testNotes · 10 questions
  9. 3.13Carrying out a two-proportion testNotes · 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. 1Comparing two groups starts with a difference of sample proportions
  2. 2The distribution is centered at the true difference in proportions
  3. 3The standard deviation combines the variability of both samples
  4. 4Large samples make the distribution approximately normal
  5. 5Four conditions must hold before using the normal model
  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

Sampling distribution of the difference

Density-0.100.10.20.30123456Difference in sample proportions
  • Normal curve
  • Mean 0.10