Tools / free calculator
Multiple comparisons calculator
Paste your p-values and see what survives once the number of tests is accounted for. Bonferroni, Holm and Benjamini-Hochberg side by side, so you can see how much of your result depends on which correction you chose.
Choosing the method that keeps your finding significant is a form of p-hacking, and a reviewer can usually tell. State the correction in the protocol or the analysis plan, then apply it whatever it does.
Have the analysis checked →In short
Which correction should you use?
Use Holm when a single false positive would be costly, for example a confirmatory trial with several primary outcomes. It controls the chance of any false positive and is uniformly better than Bonferroni, so there is no good reason to prefer plain Bonferroni. Use Benjamini-Hochberg for exploratory work with many tests, where you can tolerate a known proportion of false discoveries among the findings you follow up. Decide which before you look at the results.
Common questions
Answers before you ask
What is the difference between Bonferroni, Holm and Benjamini-Hochberg?
Bonferroni and Holm control the family-wise error rate, the chance of making any false positive at all. Benjamini-Hochberg controls the false discovery rate, the expected share of your significant findings that are false. Holm is a step-down version of Bonferroni that is always at least as powerful, so Holm should be preferred to plain Bonferroni. Benjamini-Hochberg is less strict than both and is suited to exploratory work.
Is Bonferroni too conservative?
With many tests, yes. Dividing by the number of comparisons makes it very hard to detect anything real once you pass a handful of tests. Holm gives the same protection with more power at no cost, and Benjamini-Hochberg is the usual choice when you have dozens of tests and are generating hypotheses rather than confirming one.
How many tests count as my family?
This is a judgement, not a calculation, and it should be made and written down in advance. A common approach is one family per research question: all outcomes bearing on the same question are corrected together, while genuinely separate questions are treated separately. Deciding the family after seeing results is not defensible.
Do I always need to correct?
No. A single pre-specified primary outcome needs no correction. Correction matters when several tests bear on the same conclusion, or when you are searching across many outcomes and will report whichever stands out. Exploratory analyses can be reported uncorrected provided they are labelled as exploratory and not presented as confirmatory findings.
Is my data stored?
No. Everything runs in your browser. Nothing you paste is transmitted or stored, and there is no account.
Keep reading