Multiple testing

Povilas

Hey everyone,

I'm new with Bayesian approach and my background is not statistics so I'm struggling to understand why Bayesian approach does not require correction for multiple testing? All that I can found is very technical stuff which I don't understand.

Thanks for your answers

EJ

This depends on what kind of Bayesian you are.

1. A subjective Bayesian will have a prior plausibility assignment for every possible hypothesis. Data come in, the plausibilities are updated, and that's it. No adjustment needed.
2. An objective Bayesian will change the prior model probability whenever more hypotheses are tested. Some methods to do this are described here: https://psyarxiv.com/s56mk

Cheers,

E.J.

andersony3k

Here's a relevant and recent in-press paper: https://sites.google.com/view/complexcognitionlab-bgsu/research/publications-references/lab-publications/biasing-the-input

emmacampbell

Hello @EJ ,

I've read this thesis in the hopes of understanding how to control for multiplicity in the context of my research where I do multiple independent sample t-tests. I understand the the null control method is automatically implemented in the post-hoc tests for the ANOVAs. But is there a way to appy this method in 2 sample t-tests on JASP?

Thanks,

Emma

EJ

You mean when you do multiple t-tests? This is essentially how the post-hoc ANOVA tests work: they are the result of combining a t-test Bayes factor with a correction term that depends on the number of tests that are conducted.

Cheers,

E.J.

emmacampbell

Thank you for your answer!

I found another forum channel here where the procedure is detailed by you and another user! Very helpful!

Have a good day :)

Emma