Sampling Distribution for the Difference in Sample Proportions in AP Statistics

In AP Statistics, the sampling distribution for the difference in sample proportions is the distribution of all possible values of p̂₁ − p̂₂ from two independent random samples (or randomized groups), with mean p₁ − p₂ and standard deviation √(p₁(1−p₁)/n₁ + p₂(1−p₂)/n₂).

What is the Sampling Distribution for the Difference in Sample Proportions?

Picture taking a random sample from City A and a random sample from City B, finding the proportion who exercise in each, and subtracting. Now imagine doing that over and over. You'd get a slightly different difference every time. The sampling distribution for the difference in sample proportions is the distribution of all those possible p̂₁ − p̂₂ values. It tells you what differences are typical just from random sampling, and which ones would be surprising.

Per CED essential knowledge 3.9.A.1, the distribution is centered at the true difference, μ = p₁ − p₂, so p̂₁ − p̂₂ is an unbiased estimator. Its spread is σ = √(p₁(1−p₁)/n₁ + p₂(1−p₂)/n₂). Notice the plus sign. The two samples each add their own variability, so the variances add (3.9.B lists the conditions that make this work). You need random sampling or random assignment, the 10% condition for each sample when sampling without replacement, and the large counts condition (n₁p₁, n₁(1−p₁), n₂p₂, n₂(1−p₂) all ≥ 10) before you can call the shape approximately normal.

Why the Sampling Distribution for the Difference in Sample Proportions matters in AP® Statistics

This term lives in Topic 3.9 of Unit 3 (Inference for Categorical Data: Proportions) and supports three learning objectives. AP Stats 3.9.A asks you to calculate the mean and standard deviation. AP Stats 3.9.B asks you to justify the conditions. AP Stats 3.9.C asks you to interpret the mean, standard deviation, and probabilities in the context of two specific populations. It matters because it's the engine under every two-proportion comparison you'll do. The two-sample z-interval and z-test for p₁ − p₂ both assume this distribution is approximately normal, so if you can't justify it, the inference falls apart. For the full topic walkthrough, see the 3.9 study guide.

How the Sampling Distribution for the Difference in Sample Proportions connects across the course

Sample Proportion (Unit 3)

This is the one-sample sampling distribution of p̂, doubled up. Each p̂ has its own mean p and variance p(1−p)/n. The difference distribution just stacks two of them, subtracting the means and adding the variances. If you've got the single-proportion version down, you already know 80% of this.

Combining Random Variables (Unit 2)

The 'variances add even when you subtract' rule comes straight from Unit 2. For independent random variables, Var(X − Y) = Var(X) + Var(Y). That's exactly why the standard deviation formula has a plus sign and why independence of the two samples is a required condition.

Random Assignment in Experiments (Unit 1)

Per 3.9.B.2, data from an experiment only need the randomization condition, met by randomly assigning treatments to experimental units. The 10% condition doesn't apply because you aren't sampling from a population. So your Unit 1 design knowledge decides which conditions you have to check here.

Standard Error and Two-Proportion Inference (Unit 3)

In real studies you don't know p₁ and p₂, so you plug in p̂₁ and p̂₂ and call the result the standard error. That standard error becomes the denominator of the two-proportion z-statistic and sets the width of the confidence interval for p₁ − p₂.

Is the Sampling Distribution for the Difference in Sample Proportions on the AP® Statistics exam?

Multiple-choice questions usually hand you two groups and ask you to do one of four things. You might find the mean of p̂₁ − p̂₂, compute its standard deviation, find a probability (like the chance the sample difference exceeds some value), or spot when the distribution would NOT be approximately normal. A typical stem compares adults who exercise regularly in City A versus City B, with samples like 200 and 250 people. A conditions question might give several scenarios where one has a group with too few expected successes or failures. That one fails large counts, so its distribution isn't approximately normal.

Here's a quick worked example. Say p₁ = 0.60 with n₁ = 200 and p₂ = 0.50 with n₂ = 250. The mean is 0.10, and the SD is √(0.6·0.4/200 + 0.5·0.5/250) ≈ 0.047. The counts are 120, 80, 125, and 125, all ≥ 10, so normal is fine.

No released FRQ has used this exact phrase, but the skill shows up constantly inside two-proportion inference FRQs. There, you have to name the conditions (two independent random samples or random assignment, 10% for each sample, large counts for each group) and interpret results in context. Always name both populations, like "the difference in the proportion of adults who exercise in City A and City B."

The Sampling Distribution for the Difference in Sample Proportions vs Standard Error

The standard deviation of the sampling distribution uses the true parameters p₁ and p₂. It describes how p̂₁ − p̂₂ actually varies. The standard error uses the sample statistics p̂₁ and p̂₂ as stand-ins, because in practice you don't know the parameters. Topic 3.9 problems often give you p₁ and p₂ directly, so you compute σ. Once you're doing inference with real data, you're computing a standard error.

Key things to remember about the Sampling Distribution for the Difference in Sample Proportions

• The sampling distribution of p̂₁ − p̂₂ is centered at the true difference p₁ − p₂, so the difference in sample proportions is an unbiased estimator.

• The standard deviation is √(p₁(1−p₁)/n₁ + p₂(1−p₂)/n₂), and you add the two variances even though you subtract the proportions.

• For two independent random samples, you need randomness, the 10% condition for each sample, and large counts for both groups.

• If the data come from an experiment, random assignment of treatments meets the randomization condition, and the 10% condition isn't needed.

• The distribution is approximately normal only when n₁p₁, n₁(1−p₁), n₂p₂, and n₂(1−p₂) are all at least 10.

• Every interpretation of the mean, standard deviation, or a probability has to name the two specific populations being compared.

Frequently asked questions about the Sampling Distribution for the Difference in Sample Proportions

What is the sampling distribution for the difference in sample proportions in AP Stats?

It's the distribution of all possible values of p̂₁ − p̂₂ you'd get from repeated independent samples of two groups. Its mean is p₁ − p₂ and its standard deviation is √(p₁(1−p₁)/n₁ + p₂(1−p₂)/n₂), per CED Topic 3.9.

Do you subtract the standard deviations when finding the SD of p̂₁ − p̂₂?

No. You add the variances, p₁(1−p₁)/n₁ + p₂(1−p₂)/n₂, and then take the square root. Each sample brings its own random variability, so subtracting the proportions makes the difference more variable, not less.

Do you need the 10% condition for a two-proportion experiment?

No. Per 3.9.B.2, experimental data only need the randomization condition, which you meet by randomly assigning treatments to experimental units. The 10% condition applies when you're sampling without replacement from two populations.

When is the sampling distribution of p̂₁ − p̂₂ not approximately normal?

It isn't approximately normal when any of the four expected counts, n₁p₁, n₁(1−p₁), n₂p₂, or n₂(1−p₂), is below 10. For example, a group of 50 with p = 0.10 has only 5 expected successes, so the large counts condition fails.

How is the difference in sample proportions different from a single sample proportion?

A single p̂ estimates one population proportion, with mean p and SD √(p(1−p)/n). The difference p̂₁ − p̂₂ compares two groups, so its mean is p₁ − p₂, its variance adds two pieces, and you check conditions for both samples separately.