Hypotheses for Paired t-Tests
You'll learn: How to formally state null and alternative hypotheses when testing the mean difference in paired data.
When you're working with paired data—like before-and-after measurements on the same subjects—you need to frame your research question as a pair of competing hypotheses about the mean of the differences.
The Core Idea
Remember that in a paired t-test, you first calculate the difference for each pair (d = measurement₂ - measurement₁). Your hypotheses make claims about μ_d, the true population mean of these differences.
The Null Hypothesis (H₀)
The null hypothesis typically states that there's no systematic difference between the paired measurements:
H₀: μ_d = 0
This means "on average, the difference is zero—the treatment/time/condition had no effect."
The Alternative Hypothesis (H₁)
The alternative depends on your research question:
- Two-sided: H₁: μ_d ≠ 0 (there is some difference, positive or negative)
- One-sided (greater): H₁: μ_d > 0 (the second measurement tends to be higher)
- One-sided (less): H₁: μ_d < 0 (the second measurement tends to be lower)
Example in Context
Suppose you're testing whether a training program improves test scores. You measure each student before and after training:
- H₀: μ_d = 0 (training has no effect on average score change)
- H₁: μ_d > 0 (training increases scores on average)
Here, d = (after score) - (before score), and you expect positive differences if training helps.
Key Takeaway: For paired t-tests, always formulate hypotheses about μ_d (the mean difference), not about the individual measurements themselves—the pairing is what makes this test powerful.