Confidence Interval for Paired Data
What you'll learn: How to construct confidence intervals when your data comes in natural pairs—like before/after measurements or matched subjects.
Why Paired Data is Different
Sometimes observations aren't independent samples from two different populations. Instead, they come in natural pairs: the same person measured before and after treatment, twins matched in a study, or the same plot of land tested in two seasons.
The key insight: instead of comparing two separate groups, we analyze the differences within each pair. This controls for individual variation and often gives us more precision.
The Setup
Suppose you measure n pairs: (X₁, Y₁), (X₂, Y₂), ..., (Xₙ, Yₙ).
Define the difference for each pair: Dᵢ = Xᵢ - Yᵢ
Now you have a single sample of differences: D₁, D₂, ..., Dₙ. The population mean difference μ_D is what we want to estimate.
Constructing the Paired t-Interval
Since we're working with a single sample of differences, this becomes a one-sample problem:
- Calculate the sample mean of differences: d̄ = (Σ Dᵢ)/n
- Calculate the sample standard deviation of differences: s_D
- Use the t-distribution with n - 1 degrees of freedom (because we have n pairs, not 2n independent observations)
The (1 - α)100% confidence interval is:
d̄ ± t*_(n-1) · (s_D/√n)
where t*_(n-1) is the critical value from the t-distribution.
When to Use It
- Before/after studies (weight loss, blood pressure reduction)
- Matched pairs (twins, siblings, matched controls)
- Repeated measures on the same subject
- Any situation where observations come in natural, dependent pairs
Key Takeaway: For paired data, calculate the difference within each pair first, then construct a confidence interval for the mean difference using the one-sample t-interval formula—this accounts for the dependency structure and usually provides greater statistical power than treating the data as independent samples.