Confidence Intervals for Paired Differences
You'll learn how to build confidence intervals when comparing the same subjects measured twice, like before-and-after studies.
The Core Idea
Sometimes we measure the same individuals under two conditions—like patients' blood pressure before and after medication, or students' test scores before and after tutoring. Instead of treating these as two independent groups, we calculate the difference for each pair and build a confidence interval for the mean difference.
This is called the paired t-interval because we're working with differences within pairs, not separate groups.
Why Paired Data is Different
When measurements come from the same subjects, they're correlated. A tall person might have high readings in both conditions; a short person might have low readings in both. By focusing on differences, we remove individual variation and get more precise estimates.
Think of it like weighing yourself before and after vacation—you care about the change in your weight, not two separate weights from two different people.
The Paired t-Interval Formula
First, calculate each pair's difference: d = (measurement₂ - measurement₁)
Then construct the interval:
CI = d̄ ± t* × (sd / √n)
Where:
- d̄ = mean of all differences
- sd = standard deviation of differences
- n = number of pairs
- t* = critical value from t-distribution with (n-1) degrees of freedom
Notice this looks just like a one-sample confidence interval—because that's exactly what it is! We've transformed paired data into a single sample of differences.
Key Takeaway: For paired or matched samples, calculate the difference for each pair and construct a confidence interval for the mean difference using the paired t-interval, treating differences as a single sample.