Bonferroni cutoff for eight tests

Question

A researcher runs 8 hypothesis tests and wants the family-wise error rate (the chance of any false positive across the 8) held at 5%.

Using the Bonferroni correction, what per-test p-value cutoff must each test beat?

Accepted Answer

Divide the budget. Bonferroni spreads the 5% total false-alarm budget evenly over the tests. With m tests and target \alpha, each test must clear \frac{\alpha}{m} = \frac{0.05}{8} = 0.00625. So a result is significant only if its p-value is at most 0.00625. Why it works. Each test now risks at most 0.625% of a false alarm. Summing the worst case over all 8 (the union bound) gives at most 8 	imes 0.00625 = 0.05, so the family-wide false-alarm chance stays within 5%. Sanity check. A p-value of 0.02 would pass the naive 0.05 test but fails the Bonferroni bar of 0.00625, so it is not declared si…

Bonferroni cutoff for eight tests

Your answer

More Statistics questions