01 / 21
Δ

Sampling Distributions
Differences in Sample Means

Explore how the distribution of \(\bar{X}_1 - \bar{X}_2\) behaves, what determines its center and spread, and how to use it for inference about two populations.

AP Statistics

Why Compare Two Means?

In statistics, we often want to compare two populations. For example: Is the average blood pressure lower for a treatment group than a control group? Does one fertilizer produce a higher mean yield than another?

The natural estimator for the difference in population means \(\mu_1 - \mu_2\) is the difference in sample means \(\bar{X}_1 - \bar{X}_2\). But a single sample gives only one value: to draw conclusions, we need to understand the entire sampling distribution of that difference.

Visualizing the Sampling Distribution

Pop 1 Pop 2 difference Sample Mean Values Frequency

Three Building Blocks

Center
The mean of \(\bar{X}_1 - \bar{X}_2\) equals \(\mu_1 - \mu_2\). The estimator is unbiased.
Spread
The standard deviation depends on both population variances and both sample sizes: \(\sigma_1, \sigma_2, n_1, n_2\).
Shape
By the CLT, the distribution is approximately Normal for large enough \(n_1\) and \(n_2\).

Core Formulas

1
Mean: \(\mu_{\bar{X}_1 - \bar{X}_2} = \mu_1 - \mu_2\)
2
Standard Deviation (independent samples): $$\sigma_{\bar{X}_1 - \bar{X}_2} = \sqrt{\frac{\sigma_1^2}{n_1} + \frac{\sigma_2^2}{n_2}}$$
3
Shape: If both populations are Normal, or \(n_1 \geq 30\) and \(n_2 \geq 30\), then \(\bar{X}_1 - \bar{X}_2\) is approximately Normal.

Why Variances Add

Var of Sample 1
\(\frac{\sigma_1^2}{n_1}\)
+ Var of Sample 2
\(\frac{\sigma_2^2}{n_2}\)
= Var of Difference
\(\frac{\sigma_1^2}{n_1} + \frac{\sigma_2^2}{n_2}\)

For independent random variables, variances always add: whether the operation is addition or subtraction. The property \(\text{Var}(A - B) = \text{Var}(A) + \text{Var}(B)\) holds because independence eliminates the covariance term.

Shape: The CLT Condition

The Central Limit Theorem guarantees that \(\bar{X}_1 - \bar{X}_2\) is approximately Normal when:

1
Both populations are Normal (any sample size works), or
2
Both sample sizes are large (\(n_1 \geq 30\) and \(n_2 \geq 30\)) regardless of population shape.
3
The two samples are independent of each other (drawn from separate populations or randomly assigned).

One Mean vs. Difference of Means

One Sample Mean \(\bar{X}\) Difference \(\bar{X}_1 - \bar{X}_2\)
Center \(\mu\) \(\mu_1 - \mu_2\)
Std Dev \(\dfrac{\sigma}{\sqrt{n}}\) \(\sqrt{\dfrac{\sigma_1^2}{n_1} + \dfrac{\sigma_2^2}{n_2}}\)
Shape Normal if population Normal or \(n \geq 30\) Normal if both populations Normal or both \(n \geq 30\)
Parameters \(\mu, \sigma, n\) \(\mu_1, \mu_2, \sigma_1, \sigma_2, n_1, n_2\)

Step-by-Step Application

1
Identify the populations: What are Population 1 and Population 2? Note \(\mu_1, \sigma_1, n_1\) and \(\mu_2, \sigma_2, n_2\).
2
Check conditions: Independence of the two samples? Normality or large \(n\)? 10% condition?
3
Compute the mean: \(\mu_{\bar{X}_1 - \bar{X}_2} = \mu_1 - \mu_2\).
4
Compute the standard deviation: \(\sigma_{\bar{X}_1 - \bar{X}_2} = \sqrt{\frac{\sigma_1^2}{n_1} + \frac{\sigma_2^2}{n_2}}\).
5
Find probabilities or build intervals: Use the Normal model \(N(\mu_1 - \mu_2,\; \sigma_{\bar{X}_1 - \bar{X}_2})\) and compute \(z\)-scores.

Example 1: Finding the Standard Deviation

A researcher studies test scores in two schools. School A: \(\mu_1 = 78\), \(\sigma_1 = 10\), \(n_1 = 40\). School B: \(\mu_2 = 72\), \(\sigma_2 = 12\), \(n_2 = 50\). Find the mean and standard deviation of \(\bar{X}_1 - \bar{X}_2\).

Example 2: Finding a Probability

Using the same schools (\(\mu_{\bar{X}_1-\bar{X}_2}=6\), \(\sigma=2.32\)): What is the probability that \(\bar{X}_1 - \bar{X}_2 < 2\)?

2 6 0.0427
\(N(6, 2.32)\)

There is only about a 4.3% chance that the difference in sample means falls below 2, given the true difference is 6.

Example 3: Effect of Sample Size

What happens to the standard deviation if we increase both sample sizes to \(n_1 = 100\) and \(n_2 = 100\), keeping \(\sigma_1 = 10\) and \(\sigma_2 = 12\)?

Original
\(n_1=40,\; n_2=50\)
SD \(\approx 2.32\)
Larger samples
\(n_1=100,\; n_2=100\)
SD \(\approx 1.56\)

The 10% Condition

When sampling without replacement, observations are not truly independent. The standard deviation formula works well only if each sample is small relative to its population:

✓
\(n_1 \leq 0.10 \cdot N_1\) (sample 1 is at most 10% of population 1)
✓
\(n_2 \leq 0.10 \cdot N_2\) (sample 2 is at most 10% of population 2)

The Complete Picture

Population 1
\(\mu_1, \sigma_1\)
Population 2
\(\mu_2, \sigma_2\)
Sample 1
\(n_1 \text{ obs} \rightarrow \bar{X}_1\)
Sample 2
\(n_2 \text{ obs} \rightarrow \bar{X}_2\)
\(\bar{X}_1 - \bar{X}_2\)
\(\text{Mean: } \mu_1 - \mu_2\)
\(\text{SD: } \sqrt{\frac{\sigma_1^2}{n_1} + \frac{\sigma_2^2}{n_2}}\)

Common Mistakes

✕ Subtracting Variances
Students write \(\sigma_1^2/n_1 - \sigma_2^2/n_2\). Variances always add for independent variables, regardless of whether you add or subtract the means.
✕ Forgetting Independence
The variance addition rule requires independent samples. If samples are paired or dependent, a completely different method is needed (matched pairs).
✕ Skipping Normality Check
The Normal model only applies if both populations are Normal or both \(n \geq 30\). With small samples from skewed populations, the approximation fails.

A listening study with fixed response cards and moving treatment badges

Participants wear headphones at matching task stations while a researcher distributes cards.Measurements taken from the same person remain linked. Compare within-person differences instead of treating the two columns as independent samples.

Keep the pairs linked

Six linked before-and-after pairs become six differences; within-ID sign swaps generate 64 assignments with 12 in the marked absolute tail.Suppose the same six IDs have differences 2, −1, 3, 0, 2 and 1, averaging 7/6. Under a sharp zero-effect null with randomized or exchangeable within-ID labels, 2⁶ = 64 sign assignments are allowed; 12 have absolute mean at least 7/6. Repeated measurement alone does not establish that exchangeability.

Summary: Key Facts

Property Formula / Rule
Mean \(\mu_1 - \mu_2\)
Standard Deviation \(\sqrt{\dfrac{\sigma_1^2}{n_1} + \dfrac{\sigma_2^2}{n_2}}\)
Shape Normal (if both pops Normal or both \(n \geq 30\))
Independence Two samples must be independent
10% Condition \(n_1 \leq 0.10 N_1\) and \(n_2 \leq 0.10 N_2\)
Effect of larger \(n\) SD decreases, distribution narrows

Question 1 of 5

The mean of the sampling distribution of \(\bar{X}_1 - \bar{X}_2\) equals:

Question 2 of 5

When computing the standard deviation of \(\bar{X}_1 - \bar{X}_2\), why do the variance terms add rather than subtract?

Question 3 of 5

Population 1 has \(\mu_1 = 50, \sigma_1 = 8, n_1 = 64\). Population 2 has \(\mu_2 = 45, \sigma_2 = 6, n_2 = 36\). Find \(\sigma_{\bar{X}_1 - \bar{X}_2}\).

Question 4 of 5

Which of the following is NOT a required condition for the sampling distribution of \(\bar{X}_1 - \bar{X}_2\) to be approximately Normal?

Question 5 of 5

If you quadruple both sample sizes (multiply each by 4), the standard deviation of \(\bar{X}_1 - \bar{X}_2\) is multiplied by: