I ran 80,000 simulations to investigate different p-value adjustments

Multiple comparisons are hard. Which adjustment best deals with both false positives and false negatives?

TL;DR:

• It is common in real-world contexts to do many significance tests at once. However, this translates into a greater likelihood of finding a false-positive relationship. There are a host of p-value adjustments that look to control for this, but they are often too conservative and have very high false-negative rates–especially when performing hundreds or thousands of tests.
• I …

data science p-value simulations statistics value