Course contentsShow
← Data Science
Lesson 1504 of 2,14537. A/B TestingFree lesson

Holm-Bonferroni Method

A more powerful stepwise procedure that adjusts p-values sequentially while still controlling FWER at the desired level.

Holm-Bonferroni Method

What you'll learn: How to adjust multiple p-values sequentially for greater statistical power while still controlling the family-wise error rate.

The Problem with Standard Bonferroni

You've learned that Bonferroni correction divides your alpha level (say, 0.05) equally among all tests. While this controls FWER effectively, it's conservative—meaning you might miss real effects because the corrected threshold is so strict. The Holm-Bonferroni method offers a smarter, more powerful alternative.

How Holm-Bonferroni Works

Instead of treating all tests identically, Holm-Bonferroni uses a stepwise procedure:

  1. Order your p-values from smallest to largest: p₁ ≤ p₂ ≤ ... ≤ pₘ
  2. Test sequentially starting with the smallest p-value
  3. For the i-th smallest p-value, compare it to α/(m - i + 1)
  4. Stop at the first p-value that fails to reject; all subsequent tests are also not rejected
  5. Reject all hypotheses up to (but not including) that stopping point

Why It's More Powerful

Think of it like a VIP line at a concert: the most promising results (smallest p-values) get the easiest entry threshold. The first test needs to beat α/m, but if you're testing 5 hypotheses and reach the 3rd test, it only needs to beat α/3—a much easier bar than Bonferroni's uniform α/5.

This sequential adjustment means you're less likely to throw away genuine discoveries while still maintaining FWER control at your chosen alpha level.

Key Takeaway: Holm-Bonferroni is a stepwise improvement over standard Bonferroni that adjusts thresholds based on the rank of each p-value, providing greater statistical power while maintaining the same FWER protection.