Course contentsShow
← Mathematics
Lesson 3093 of 3,12160. Statistics — ModelingFree lesson

Introduction to Analysis of Variance (ANOVA)

Understanding when and why to use ANOVA to compare means across multiple groups instead of multiple t-tests.

Introduction to Analysis of Variance (ANOVA)

You'll learn why ANOVA is essential when comparing means across three or more groups, and why running multiple t-tests isn't the answer.

Why Not Just Run Multiple t-Tests?

Imagine you're comparing average test scores across four different teaching methods. Your first instinct might be to run separate two-sample t-tests for every pair: Method 1 vs. 2, Method 1 vs. 3, Method 2 vs. 3, and so on. That's six different tests!

Here's the problem: each t-test carries a Type I error risk (false positive). If you set α = 0.05 for each test, you're giving yourself a 5% chance of incorrectly finding a difference in each comparison. With multiple tests, these errors compound. Your overall chance of making at least one Type I error across all six tests jumps to roughly 26%—much higher than your intended 5%.

What ANOVA Does Differently

Analysis of Variance (ANOVA) tests whether any of the group means differ, all in a single test that controls the Type I error rate at your chosen significance level. Think of it as a gatekeeper: it tells you "Yes, there's evidence that at least one group differs from the others" or "No, we can't conclude any groups differ."

ANOVA decomposes the total variation in your data into:

  • Between-group variation: differences among group means
  • Within-group variation: natural variability within each group

It uses an F-test (which you've seen before) to compare these sources of variation.

When to Use ANOVA

Use ANOVA when:

  • You're comparing three or more group means
  • You want to control your overall Type I error rate
  • Groups are independent and data meet regression-like assumptions

Key Takeaway: ANOVA lets you compare multiple group means simultaneously while controlling Type I error, avoiding the inflated error risk of running multiple pairwise t-tests.