01 / 21

Justifying Claims:
Difference of Two Means

Use confidence intervals for \(\mu_1 - \mu_2\) to determine whether there is convincing evidence of a real difference between two population means.

AP Statistics

What Is a Two-Sample Confidence Interval?

When we collect data from two independent groups, we often want to know whether their population means differ. A confidence interval for \(\mu_1 - \mu_2\) gives us a range of plausible values for that difference.

The Big Idea: Does the Interval Contain Zero?

The key question is whether zero falls inside or outside the confidence interval. If zero is not in the interval, we have convincing evidence that the two population means truly differ.

0 Excludes 0 Evidence of difference Includes 0 No convincing evidence

Why Does This Matter?

Medical Trials
Does a new drug lower blood pressure more than the current treatment?
Education
Do students taught with method A score higher than those with method B?
Business
Is average spending per customer different between two store locations?

The Confidence Interval Formula

\(\bar{x}_1 - \bar{x}_2\)
Point estimate: the observed difference in sample means.
\(t^*\)
Critical value from the \(t\)-distribution for the chosen confidence level.
\(\sqrt{\frac{s_1^2}{n_1}+\frac{s_2^2}{n_2}}\)
Standard error of the difference, combining variability from both samples.

Conditions for Valid Inference

1
Random: Both samples are collected using a random process (random sampling or random assignment).
2
Independent: The two samples are independent of each other, and each sample is \(< 10\%\) of its population (10% condition).
3
Normal: Both sample sizes are \(> 30\) (CLT), or both population distributions are approximately Normal. Check for strong skew or outliers.

Confidence Interval vs. Hypothesis Test

Confidence Interval Hypothesis Test
Goal Estimate a range of plausible values for \(\mu_1 - \mu_2\) Decide whether to reject \(H_0\!: \mu_1 = \mu_2\)
Output An interval, e.g. \((1.2,\, 4.8)\) A \(p\)-value and decision
Connection If the CI excludes 0, a two-sided test at the same \(\alpha\) rejects \(H_0\)
Advantage Shows direction and magnitude Directly tests a specific claim

How to Interpret the Interval

1
State the interval: "We are 95% confident that the true difference \(\mu_1 - \mu_2\) is between \(a\) and \(b\)."
2
Check for zero: Does the interval contain 0? If no, there is convincing evidence of a difference.
3
State direction: If both bounds are positive, \(\mu_1 > \mu_2\). If both are negative, \(\mu_1 < \mu_2\).
4
Justify in context: Connect the statistical conclusion back to the real-world scenario.

Step-by-Step: Building the Argument

1
Check conditions: Verify Random, Independent, and Normal for both samples.
2
Compute the point estimate: Calculate \(\bar{x}_1 - \bar{x}_2\).
3
Find the standard error: \(SE = \sqrt{\frac{s_1^2}{n_1} + \frac{s_2^2}{n_2}}\).
4
Find \(t^*\): Use a \(t\)-table or calculator with appropriate degrees of freedom.
5
Construct the CI: \((\bar{x}_1 - \bar{x}_2) \pm t^* \cdot SE\).
6
Interpret and justify: State whether 0 is in the interval and what that means in context.

Example 1: Sleep Study

A researcher wants to compare average hours of sleep between college students who exercise regularly (\(n_1 = 40\)) and those who do not (\(n_2 = 35\)). The data:

Exercisers
\(\bar{x}_1 = 7.4\) hrs, \(s_1 = 1.2\), \(n_1 = 40\)
Non-Exercisers
\(\bar{x}_2 = 6.7\) hrs, \(s_2 = 1.5\), \(n_2 = 35\)

Example 1: Calculation

1
Point estimate: \(\bar{x}_1 - \bar{x}_2 = 7.4 - 6.7 = 0.7\) hours.
2
Standard error: \(SE = \sqrt{\frac{1.2^2}{40} + \frac{1.5^2}{35}} = \sqrt{0.036 + 0.0643} = \sqrt{0.1003} \approx 0.317\).
3
Critical value: Using \(df \approx 64\) (calculator) at 95% confidence: \(t^* \approx 2.00\).
4
CI: \(0.7 \pm 2.00(0.317) = 0.7 \pm 0.634 = (0.066,\, 1.334)\).

Example 2: Test Scores

A school compares average SAT math scores between students who used a prep course (\(n_1 = 50\)) and those who studied independently (\(n_2 = 45\)).

Prep Course
\(\bar{x}_1 = 612\), \(s_1 = 78\), \(n_1 = 50\)
Independent
\(\bar{x}_2 = 598\), \(s_2 = 85\), \(n_2 = 45\)

Example 2: Result and Interpretation

1
SE: \(\sqrt{\frac{78^2}{50} + \frac{85^2}{45}} = \sqrt{121.68 + 160.56} = \sqrt{282.24} \approx 16.80\).
2
With \(t^* \approx 1.987\) at 95%: \(14 \pm 1.987(16.80) = 14 \pm 33.38 = (-19.38,\, 47.38)\).
0 -19.4 47.4 14

Example 3: Interpreting a Given CI

A 90% confidence interval for the difference in mean commute times (city A minus city B) is \((3.1, 11.7)\) minutes. What can we conclude?

0 3.1 11.7 7.4

Common Mistakes to Avoid

✕ Ignoring Zero
Claiming a difference exists without checking whether the interval contains 0.
✕ Wrong Interpretation
Saying "95% chance the true difference is in this interval." The true difference is fixed: it either is or isn't in the interval.
✕ Skipping Conditions
Constructing a CI without verifying Random, Independent, and Normal conditions first.

Summary: Decision Framework

CI for \(\mu_1 - \mu_2\) Conclusion Justification
Entirely positive \(\mu_1 > \mu_2\) 0 is below interval
Entirely negative \(\mu_1 < \mu_2\) 0 is above interval
Contains zero No convincing evidence 0 is a plausible value for the difference

Question 1 of 5

A 95% confidence interval for \(\mu_1 - \mu_2\) is \((2.3, 8.9)\). Which conclusion is best supported?

Question 2 of 5

A 90% CI for the difference in mean test scores (Group A minus Group B) is \((-4.2, 12.6)\). What should you conclude?

Question 3 of 5

Which of the following is the correct interpretation of a 95% confidence interval \((1.5, 6.3)\) for \(\mu_1 - \mu_2\)?

Question 4 of 5

Given: \(\bar{x}_1 = 45\), \(s_1 = 8\), \(n_1 = 36\), \(\bar{x}_2 = 40\), \(s_2 = 10\), \(n_2 = 36\). What is the standard error of \(\bar{x}_1 - \bar{x}_2\)?

Question 5 of 5

A 95% CI for \(\mu_1 - \mu_2\) is \((-12.1, -0.3)\). A student claims there is no significant difference. Is this justified?