Course contentsShow
← Data Science
Lesson 311 of 2,1459. Hypothesis Testing FoundationsFree lesson

One-Sided vs Two-Sided Alternatives

Distinguish between directional (greater/less than) and non-directional (not equal) alternative hypotheses.

One-Sided vs Two-Sided Alternatives

What you'll learn: How to choose between directional and non-directional alternative hypotheses based on what you're trying to prove.

The Core Idea

When you set up a hypothesis test, your alternative hypothesis (H₁) can point in three possible directions. You need to decide: are you testing if something is different, or specifically if it's greater or less than a benchmark?

Two-Sided (Non-Directional) Alternative

A two-sided alternative uses "not equal to" (≠). You're testing whether a parameter differs from the null value in either direction.

Example: H₀: μ = 100 vs H₁: μ ≠ 100

You don't care how it's different—just that it is. Think of it like asking, "Is this coin unfair?" without specifying whether it favors heads or tails.

One-Sided (Directional) Alternative

A one-sided alternative uses either "greater than" (>) or "less than" (<). You're testing a specific direction of change.

Example (right-tailed): H₀: μ = 100 vs H₁: μ > 100
Example (left-tailed): H₀: μ = 100 vs H₁: μ < 100

Use one-sided tests when you have a strong reason to care about only one direction. For instance, testing whether a new drug lowers blood pressure (not just changes it), or whether a process improves conversion rates.

When to Use Which?

  • Two-sided: Default choice when exploring differences without prior directional expectations
  • One-sided: When theory, cost, or practical concerns make only one direction meaningful

Key Takeaway: Two-sided alternatives test for any difference (μ ≠ value), while one-sided alternatives test for a specific direction (μ > value or μ < value). Choose based on your research question, not what makes your test easier to pass.