Multiple-Testing Correction (FDR / Bonferroni / Holm)
Running many hypothesis tests at once (multiple markers, multiple subgroups, omics) inflates false positives in the raw p-values. This tool gives four corrections at once: Bonferroni and Holm (controlling the family-wise error rate, FWER), Benjamini-Hochberg (controlling the false discovery rate, FDR) and Benjamini-Yekutieli (FDR under dependence).
① Paste p-values
One p-value per line (or separated by spaces/commas), in the range 0–1.
How to use & methodology
When is multiple-testing correction required?
When a study runs many hypothesis tests and draws 'significant or not' conclusions from them (multiple outcomes/subgroups/markers, high-throughput omics). Looking only at a single pre-specified primary endpoint usually does not need a global correction.
Bonferroni or FDR (BH)?
To strictly control 'any false positive at all', use FWER (Bonferroni/Holm); when you can tolerate a certain proportion of false positives among significant results and want higher discovery power, use FDR (BH). Exploratory, high-dimensional data usually use BH.
What's the difference between BH and BY?
BH assumes tests are independent or positively correlated; when tests may be arbitrarily correlated, BY multiplies by a Σ1/i factor for more conservative FDR control, at the cost of fewer significant results.
Can the corrected p (q-value) be reported directly?
Yes. The table gives each method's corrected p/q value; compare it with α to judge significance. This is more intuitive and verifiable than reporting only a 'corrected threshold'.