Directionality in Hypothesis Testing
What You'll Learn
You'll learn when to predict a specific direction of change versus simply testing for any difference from the null hypothesis.
The Core Idea
When setting up hypothesis tests, your research question determines whether you're looking for:
- Any difference (two-sided/non-directional)
- A specific direction (one-sided/directional)
Think of it like a detective's suspicion: If you suspect a new drug has some effect (could help or harm), that's non-directional. But if you specifically believe it improves outcomes, that's directional.
Two-Sided (Non-Directional) Tests
These test whether a parameter differs from the null in either direction:
- H₀: μ = 100
- H₁: μ ≠ 100 (could be greater or less)
Use when: Your research question is "Is there a difference?" without predicting which way.
One-Sided (Directional) Tests
These test for change in a specific direction:
- H₀: μ ≤ 100
- H₁: μ > 100 (only interested in increases)
Use when: Theory or prior evidence suggests a particular direction, and you're only interested in detecting effects in that direction.
Why It Matters
Directionality affects:
- Where your rejection region sits (both tails vs. one tail)
- Your critical values
- How you calculate p-values
- Your test's power to detect effects in the predicted direction
The choice must be made before seeing your data, based on your research question—not after peeking at results.
Example Context
Testing a new teaching method:
- Two-sided: "Does this method change test scores?" (could improve or worsen)
- One-sided: "Does this method improve test scores?" (only care about increases)
Key Takeaway: Directionality stems from your research question—use two-sided tests for detecting any difference, and one-sided tests only when theory justifies looking in a specific direction before data collection.