Justifying Claims with
Confidence Intervals
Learn how to use a confidence interval for \(p_1 - p_2\) to determine whether there is convincing evidence of a difference between two population proportions.
What Is a Confidence Interval?
A confidence interval provides a range of plausible values for an unknown population parameter. Rather than a single estimate, it communicates both the estimate and the uncertainty around it.
The Difference of Two Proportions
When comparing two populations, we often want to know whether the proportions \(p_1\) and \(p_2\) differ. The parameter of interest is \(p_1 - p_2\).
Why Does This Matter?
In real-world decisions: does a new drug work better than a placebo? Do voters in two states differ on a policy? We need a rigorous way to justify claims about differences.
The Confidence Interval Formula
Visualizing the Interval
A confidence interval is a segment on the number line. The key question: does the interval contain 0?
Contains 0 vs. Does Not
| Interval contains 0 | Interval does NOT contain 0 | |
|---|---|---|
| Meaning | 0 is a plausible value for \(p_1 - p_2\) | 0 is NOT a plausible value for \(p_1 - p_2\) |
| Conclusion | No convincing evidence that \(p_1 \neq p_2\) | Convincing evidence that \(p_1 \neq p_2\) |
| Example | (-0.05, 0.11) | (0.03, 0.15) |
| Claim | Cannot justify a claim of a difference | Can justify: the proportions are different |
Checking the Conditions
Before constructing the interval, three conditions must be verified:
Step-by-Step Approach
Example 1: Drug Trial
In a clinical trial, 120 of 400 patients on Drug A recovered, and 90 of 350 on Drug B recovered. Construct a 95% confidence interval for \(p_A - p_B\). (Assume conditions for inference have been met.)
Example 1: Interpretation
Example 2: Voter Survey
A poll finds 280 of 500 urban voters and 195 of 500 rural voters support a policy. Build a 95% CI for \(p_U - p_R\).
Example 3: Borderline Case
A study reports a 90% confidence interval for \(p_1 - p_2\) as (0.002, 0.098). A colleague claims this means the proportions are "practically the same." Is that justified?
Reading the Direction
The sign of the interval's bounds tells you which proportion is likely larger:
Common Mistakes
Summary Reference
| Step | What to Do | Key Detail |
|---|---|---|
| 1. Parameter | Define \(p_1 - p_2\) | Use context |
| 2. Conditions | Random, 10%, Large Counts | All four count checks |
| 3. Calculate | \((\hat{p}_1 - \hat{p}_2) \pm z^* \cdot SE\) | \(z^* = 1.96\) for 95% |
| 4. Interpret | "We are C% confident..." | State in context |
| 5. Justify | Does interval contain 0? | Direction if it doesn't |
Question 1 of 5
A 95% confidence interval for \(p_1 - p_2\) is (-0.07, 0.15). What can we conclude?
Question 2 of 5
A 99% CI for \(p_1 - p_2\) is (0.03, 0.21). Which claim is justified?
Question 3 of 5
Which condition requires that \(n_1\hat{p}_1 \geq 10\), \(n_1(1-\hat{p}_1) \geq 10\), \(n_2\hat{p}_2 \geq 10\), and \(n_2(1-\hat{p}_2) \geq 10\)?
Question 4 of 5
A 95% CI for \(p_1 - p_2\) is (-0.18, -0.02). A student says: "There is a 95% probability the true difference is in this interval." Is the student correct?
Question 5 of 5
Suppose 150 of 300 men and 180 of 300 women prefer Brand X. The 95% CI for \(p_M - p_W\) is approximately (-0.18, -0.02). What is the best conclusion?