# Sampling Distributions for the Difference Between Two Sample Means

> AP Statistics · AP Stats 2024-2026
> Source: https://www.owlsprep.com/study/ap-statistics-u13-sampling-distributions-for-the-difference/

This module breaks down the center, spread, and shape properties of the sampling distribution for the difference between two independent sample means, including condition checks and common AP exam calculations.

**Prerequisites:** [Single-sample sampling distributions for means](https://www.owlsprep.com/study/ap-statistics-u12-sampling-distributions-single-mean/); [Descriptive statistics for quantitative data](https://www.owlsprep.com/study/ap-statistics-u1-descriptive-stats-quantitative/)

## Learning objectives

- Define the sampling distribution of $\bar{x}_1 - \bar{x}_2$ for independent random samples
- Derive and state the mean, standard deviation, and shape conditions for this distribution
- Verify all required Normal/Large Sample and independence conditions for two-sample procedures
- Calculate probabilities for observed values of the difference in independent sample means

## Core Definition and Parameters of the Sampling Distribution

**Sampling Distribution of $\bar{x}_1 - \bar{x}_2$** — The distribution of all possible values of the difference between sample means from two independent random samples of fixed sizes $n_1$ and $n_2$ drawn from two distinct populations.

*Notation:* $\bar{x}_1 - \bar{x}_2$

*Example:* If you take 50 samples of size 10 from Population A and 50 samples of size 12 from Population B, compute $\bar{x}_A - \bar{x}_B$ for each pair, the resulting histogram approximates this distribution.

$$\mu_{\bar{x}_1 - \bar{x}_2} = \mu_1 - \mu_2$$

This result holds for any two independent samples, no matter the population distribution shape, as long as the samples are randomly selected from their respective populations.

$$\sigma_{\bar{x}_1 - \bar{x}_2} = \sqrt{\frac{\sigma_1^2}{n_1} + \frac{\sigma_2^2}{n_2}}$$

> **Critical Independence Rule**
>
> This spread formula only applies if the two samples are fully independent. If samples are paired or dependent, you cannot use this formula at all.

**Worked example:** Population 1 has mean 22 and standard deviation 5. Population 2 has mean 18 and standard deviation 3. For independent samples of size $n_1=16$ and $n_2=25$, calculate the mean and standard deviation of $\bar{x}_1 - \bar{x}_2$.

1. Calculate the mean of the sampling distribution:
2. $$\mu_{\bar{x}_1 - \bar{x}_2} = 22 - 18 = 4$$
3. Calculate the standard deviation of the sampling distribution:
4. $$\sigma_{\bar{x}_1 - \bar{x}_2} = \sqrt{\frac{5^2}{16} + \frac{3^2}{25}} = \sqrt{1.5625 + 0.36} \approx 1.39$$

> **Exam tip:** AP graders will deduct points if you incorrectly write the spread formula as $\frac{\sigma_1 - \sigma_2}{n_1 + n_2}$ — always square each population standard deviation first.

## Shape Conditions for the Normal Approximation

For the sampling distribution of $\bar{x}_1 - \bar{x}_2$ to be approximately Normal, you must satisfy three core conditions, none of which can be skipped for full AP exam credit.

- **Random**: Both samples are selected via simple random sampling from their respective populations
- **Independence**: Observations within each sample are independent, and the two samples are independent of each other; verify the 10% condition if sampling without replacement
- **Normal/Large Sample**: Either both populations are Normally distributed, or both sample sizes are ≥ 30, with no extreme outliers in small samples for roughly symmetric populations

**Check your understanding**

Test your understanding of condition checks:

1. We take a sample of size 15 from a right-skewed population and a sample of size 18 from a left-skewed population. Can we use a Normal approximation for $\bar{x}_1 - \bar{x}_2$?

   - Yes, because the total sample size is 33 which is over 30
   - No, neither sample size is ≥ 30, and the populations are skewed
   - Yes, the skews cancel out so the distribution is Normal

   *Why:* The CLT applies to each sample individually, not the sum of sample sizes, so you cannot use the Normal approximation here.

**Worked example:** We have two independent SRS samples: $n_1=40$ from Population 1 with $\sigma_1=8$, $n_2=35$ from Population 2 with $\sigma_2=6$. Both population distributions are unknown. Confirm if the sampling distribution of $\bar{x}_1 - \bar{x}_2$ is approximately Normal.

1. Check the Random condition: The problem explicitly states both are independent simple random samples, so this is satisfied.
2. Check the 10% condition: We can assume the two populations are at least 400 and 350 members respectively, standard for AP problems unless stated otherwise.
3. Check the Normal/Large Sample condition: $n_1=40 ≥30$, $n_2=35 ≥30$, so the CLT applies to both individual sampling distributions, making the difference distribution approximately Normal.

> **Exam tip:** You must explicitly name all three conditions in AP free response to earn full credit for two-sample inference procedures.

## Calculating Probabilities for $\bar{x}_1 - \bar{x}_2$

Once you confirm the sampling distribution is approximately Normal, you can standardize any observed difference $\bar{x}_1 - \bar{x}_2$ to a z-score, then use standard Normal tables or a calculator to find the corresponding probability.

$$z = \frac{(\bar{x}_1 - \bar{x}_2) - (\mu_1 - \mu_2)}{\sqrt{\frac{\sigma_1^2}{n_1} + \frac{\sigma_2^2}{n_2}}}$$

**Worked example:** Given the sampling distribution from the earlier example where $\mu_{\bar{x}_1 - \bar{x}_2}=4$ and $\sigma_{\bar{x}_1 - \bar{x}_2}=1.39$, find the probability that $\bar{x}_1 - \bar{x}_2$ is greater than 6.

1. Compute the z-score for the value 6:
2. $$z = \frac{6 - 4}{1.39} \approx 1.44$$
3. Find the area to the right of z=1.44 on the standard Normal curve.
4. This area equals $1 - \Phi(1.44) \approx 1 - 0.9251 = 0.0749$, so the probability is ~7.5%.

**Exam command terms**

Common AP exam command terms for this topic include:

- **Verify the conditions** — Explicitly state and justify all three random, independent, normal conditions, do not skip any step

- **Find the probability** — State the distribution you are using first, then show your z-score calculation, then give the final probability.

## Common pitfalls

- **Wrong:** Using the sum of standard deviations instead of the sum of variances under the square root for the spread of the difference
  - Why it fails: Variances of independent variables add, but standard deviations do not, so this will overestimate the spread.
  - Correct: Square each population standard deviation, divide by its sample size, sum those values, then take the square root of the total.
- **Wrong:** Applying the Central Limit Theorem to the total combined sample size $n_1 + n_2$ instead of checking each sample size individually
  - Why it fails: The CLT applies to each individual sampling distribution of $\bar{x}_1$ and $\bar{x}_2$, not their sum.
  - Correct: Confirm $n_1 ≥ 30$ AND $n_2 ≥ 30$ for the Normal approximation to hold for the difference.
- **Wrong:** Using sampling distribution rules for paired data on two independent samples
  - Why it fails: Paired data reduces to a single sample of differences, which uses a completely different sampling distribution.
  - Correct: Always confirm if samples are independent or paired before selecting the correct sampling distribution framework.
- **Wrong:** Calculating the mean of the difference as $\mu_1 + \mu_2$ instead of $\mu_1 - \mu_2$
  - Why it fails: The difference of two random variables has a mean equal to the difference of their individual means, not the sum.
  - Correct: Subtract the second population mean from the first to get the center of the sampling distribution.
- **Wrong:** Forgetting to apply the 10% condition when sampling without replacement for two groups
  - Why it fails: The standard deviation formula assumes observations are independent, which fails if you sample more than 10% of a small population.
  - Correct: Explicitly state that both populations are at least 10 times their respective sample sizes to satisfy independence.

## Cheatsheet

| Property | Formula / Rule |
| --- | --- |
| Mean of sampling distribution | $\mu_{\bar{x}_1 - \bar{x}_2} = \mu_1 - \mu_2$ |
| Standard deviation (independent samples) | $\sigma_{\bar{x}_1 - \bar{x}_2} = \sqrt{\frac{\sigma_1^2}{n_1} + \frac{\sigma_2^2}{n_2}}$ |
| Normal condition 1 | Both populations are Normally distributed |
| Normal condition 2 | $n_1 \geq 30$ AND $n_2 \geq 30$ |
| Z-score for difference | $z = \frac{(\bar{x}_1 - \bar{x}_2) - (\mu_1 - \mu_2)}{\sigma_{\bar{x}_1 - \bar{x}_2}}$ |

## What's next

Mastering this sampling distribution is the critical foundation for all formal two-sample t inference procedures you will encounter on the AP Statistics exam. The rules you learned here directly extend to two-sample t confidence intervals and two-sample t significance tests, which make up over 15% of the points on the AP Stats free response section most years. You will also use this framework to distinguish between paired t procedures and two-sample t procedures, one of the most commonly tested high-difficulty discrimination points on the exam. Ensure you can quickly state all three conditions for the Normal approximation, as graders will deduct points for missing condition checks even if your final numerical answer is correct.

---

From [OwlsPrep](https://www.owlsprep.com) — free study guides for A-Level, IB, AP and IGCSE, written against the official syllabus. Canonical page: https://www.owlsprep.com/study/ap-statistics-u13-sampling-distributions-for-the-difference/