Course contentsShow
← Data Science
Lesson 350 of 2,1459. Hypothesis Testing FoundationsFree lesson

Choosing the Right Tail Configuration

Learn decision criteria for selecting one-tailed versus two-tailed tests based on research context and pre-registration.

Choosing the Right Tail Configuration

What you'll learn: How to decide whether to use a one-tailed or two-tailed hypothesis test based on your research question and ethical considerations.


When Should You Choose Each Type?

Now that you understand the mechanics of one-tailed and two-tailed tests, the crucial question is: which should you use? This decision must be made before collecting or analyzing data—ideally during study pre-registration.

Use a Two-Tailed Test When:

  • You care about differences in either direction. Example: Testing whether a new drug affects blood pressure (could increase or decrease it).
  • You have no strong prior directional hypothesis. You're exploring whether something differs from a baseline.
  • Ethical or safety concerns exist. In medical trials, you must detect harmful effects, not just beneficial ones.
  • This is your default. When in doubt, use two-tailed—it's more conservative and widely accepted.

Use a One-Tailed Test When:

  • You have a specific directional prediction supported by theory or prior research.
  • Only one direction matters practically. Example: Testing if a new process is faster than the old one (slower would mean you just don't adopt it).
  • You pre-registered this directional hypothesis before seeing the data.

The Pre-Registration Principle

Never choose your tail configuration after seeing your results. This is "p-hacking" and inflates false positive rates. Think of it like placing a bet after the race ends—it defeats the purpose of the test's error control.

Analogy: Choosing tails is like setting your GPS before driving. You decide where you're looking for evidence (everywhere, or just north) before the journey, not after you've arrived somewhere convenient.


Key Takeaway: Choose two-tailed tests as your default unless you have strong, pre-specified reasons to predict a specific direction—and always document this choice before analysis to maintain scientific integrity.