Multiplicity Testing

Question

[jth]

Hi all,

I have a question on selecting a p-value correction method. I always remember Perneger's article [1] about bonferroni correction when i think of applying a correction on my p-values. To remind that, Perneger [1] said: "Bonferroni adjustments are, at best, unnecessary and, at worst, deleterious to sound statistical inference". So, i always tend to perform FDR correction. But, i'm not quite sure how to select among various correction methods. When to use what, and why? I couldn't be able to find out a solid paper describes the use cases, and comparisons of these methods.

What do you suggest?

Thanks a lot!

[1] Perneger T. V. (1998) "What's wrong with Bonferroni adjustments". BMJ 316: 1236, doi: http://dx.doi.org/10.1136/bmj.316.7139.1236

[David Westergaard]

That's a pretty delicate question, I'd say. Benjamini-Hocberg is the least stringent of the methods, and seems to be the most widely used and generally accepted. It's what I usually use for p-value correction, though I'm trying to move away from frequential hypothesis tests and into Bayesian parameter estimation.