Multiple Testing Adjustment of a P-value List
Multiple testing adjustment of a p-value list converts many raw p-values into adjusted p-values or q-values that control the family-wise error rate or the false discovery rate.
p값 목록 다중검정 보정은 많은 원 p값을 가족별 오류율이나 거짓발견율을 통제하는 보정 p값 또는 q값으로 바꾸는 절차입니다.
I ran 20,000 gene-level tests; how many results remain significant after correcting for multiple testing?
유전자 2만 개를 각각 검정했는데, 다중검정 보정을 하면 몇 개가 유의하게 남나요?
Supply the raw p-values as one column and choose a procedure: Benjamini-Hochberg, Benjamini-Yekutieli or Storey's q-value control the expected share of false discoveries, while Holm, Hochberg, Hommel, Bonferroni and Šidák control the chance of any false positive. The output lists adjusted values, the number of discoveries, the estimated proportion of true nulls and the genomic inflation factor.
When to use Multiple Testing Adjustment of a P-value List
- You have a long list of p-values from genes, SNPs, survey items or many separate tests.
- You want to control the false discovery rate in an exploratory screen.
- You need to check whether test statistics are inflated overall (λ_GC).
When to use something else
- You are comparing group means pairwise from raw data; use a post hoc test such as Tukey HSD.
- Tests are few and must each hold a strict error rate; use Holm or Bonferroni on that small set.
Assumptions and requirements
- Each p-value is valid (uniform under its null hypothesis).
- BH and Storey assume independence or positive dependence among tests; BY holds under any dependence.
- The list includes all tests that were run, not only the interesting ones.
What the analysis reports
- Adjusted p-value or q-value for the 60 smallest p-values
- Number of discoveries at the chosen level
- Storey estimate of the proportion of true null hypotheses (π0)
- Genomic inflation factor λ_GC
- Histogram of raw p-values
How to report it
Of [m] tests, [value] had unadjusted p < .05; after [method] adjustment controlling the FDR at .05, [value] remained significant (π0 = [value]; λ_GC = [value]).
Related methods
Run Multiple Testing Adjustment of a P-value List on Quorant
- Data you need
- a table (CSV or Excel) with an outcome variable
- Price
- 250 credits for one analysis up to 10,000 rows (complexity level 2); larger or wider data costs more. Optional methodologist review: 1200 credits.
- Checking
- Checked by property tests in every deployment
- Before you pay
- A free pre-analysis diagnosis checks your data and gives an exact quote; nothing is deducted until you approve it.