I'm comparing protein abundance in different disease/symptomatic states using R and correcting for multiple testing using the built-in BH method like this:
SymptomStatus_p.value = map2_dbl(model, AptName, ~ {summary(.x)$coefficients['SymptomGroup1', "Pr(>|t|)"] }), SymptomStatusfdr = p.adjust(SymptomStatus_p.value, method = "BH")
However, some of my analyses are yielding identical adjusted p-values, sometimes of 10+ proteins/genes with the same adjusted p-value. Why is this happening? Is it a problem with my code? Or is this normal for the BH correction method and if so, what causes this "flattening" of p-values to occur?
What you are seeing is normal for the BH procedure. For example for the p-values you listed:
pValue = c(1.12e-5, 3.51e-5, 0.000169, 0.000198, 0.000227, 0.000428, 0.000604, 0.000674, 0.000682)
p.adjust(pValue, method="BH")
### 0.00010080 0.00015795 0.00040860 0.00040860 0.00040860 0.00064200 0.00068200 0.00068200 0.00068200
You can read how the adjustment is done, and see another example here: https://stats.libretexts.org/Bookshelves/Applied_Statistics/Biological_Statistics_(McDonald)/06%3A_Multiple_Tests/6.01%3A_Multiple_Comparisons
- Badges
- Science topic