Bonferroni Inequalities — Definition, Formula & Example
The Bonferroni inequalities are a family of bounds on the probability that at least one of several events happens. You stop the inclusion-exclusion formula early, and the sum you have so far is guaranteed to be too high or too low depending on where you stopped.
Let be events and define . For each with , the partial sum is an upper bound for when is odd and a lower bound when is even. When the partial sum equals the probability exactly.
Key Formula
Where:
- = The events, for i = 1 to n
- = Sum of the probabilities of all intersections of exactly j distinct events
- = Number of inclusion-exclusion terms kept (even k gives the lower bound shown)
- = Probability that at least one event occurs
Worked Example
Problem: You plan a three-day camping trip. The probability of rain on any one given day is 0.2. The probability of rain on any two given days is 0.05. The probability of rain on all three days is 0.01. Use the Bonferroni inequalities to bound the probability that it rains on at least one day.
Step 1: Compute the three sums. There are 3 single days, 3 pairs of days, and 1 triple.
Step 2: Keeping one term (odd) gives an upper bound. Keeping two terms (even) gives a lower bound.
Step 3: Keeping three terms (odd) gives a sharper upper bound. Because there are only three events, this third partial sum is the exact value.
Answer: The bounds give . With three events, the full inclusion-exclusion sum shows the probability is exactly 0.46.
Why It Matters
The first Bonferroni inequality, , is the basis of the Bonferroni correction in statistics. If you run hypothesis tests, each at significance level , the chance of at least one false positive is at most . This correction appears in medical trials, genetics studies, and A/B testing.
Common Mistakes
Mistake: Treating as an upper bound, or assuming every partial sum overestimates the probability.
Correction: The direction alternates. An odd number of terms gives an upper bound, and an even number gives a lower bound. In the camping example, falls below the true value 0.46.
