Course contentsShow
← Data Science
Lesson 369 of 2,14510. Hypothesis Tests for MeansFree lesson

When to Use a Paired t-Test

Understand paired data scenarios including before-after, matched pairs, and repeated measures designs.

When to Use a Paired t-Test

You'll learn when paired data requires a special kind of t-test that accounts for natural connections between observations.

The Core Idea

A paired t-test is used when your data comes in naturally connected pairs—each observation in one group has a specific matching partner in the other group. Instead of treating two samples as independent, you analyze the differences within each pair.

Think of it like rating a movie before and after reading the book: each person gives two ratings, and those ratings are connected through the same individual. You wouldn't compare "before" people to different "after" people—you'd look at how each person changed.

Three Common Paired Data Scenarios

1. Before-After Measurements

The same subjects measured twice: blood pressure before and after medication, test scores before and after tutoring, or website speed before and after optimization.

2. Matched Pairs

Different subjects deliberately paired by characteristics: twins in medical studies, similar stores in A/B tests, or plots of land matched by soil quality in agricultural experiments.

3. Repeated Measures

The same subject measured under different conditions: reaction time with and without caffeine, typing speed on two different keyboards, or pain ratings with two treatments.

Why It Matters

Using a paired t-test when you have paired data is much more powerful than an independent two-sample t-test because it controls for subject-to-subject variability. You're essentially making each pair its own mini-experiment.

Key Takeaway: Use a paired t-test when observations naturally occur in connected pairs (same subject measured twice, or deliberately matched subjects), and analyze the differences within pairs rather than comparing groups independently.