phucvu-nyu/post-hoc-selection

This is an ongoing research project that I am working with Dr. Rebecca Betensky at NYU

The situation:

• Clinical trials are expensive, there are many outcomes studied.
• Besides the important ones or "primary outcomes", there are secondary and exploratory outcomes that can be used to generate new hypotheses
• The New England Journal of Medicine requires investigators to prespecify a multiplicity adjustment method for secondary and exploratory outcomes if they want to report the p-values

What we argue:

• The purpose of secondary outcomes is just to generate new hypotheses
• A post-hoc approach for multiplicity adjustment method may be beneficial and increase the power significantly

What's here?

• The code in R for my simulation to calculate the FWER and FDR of the post hoc approach
• The code in Python for data visualization