1 Comment
User's avatar
⭠ Return to thread
Frank Harrell's avatar

Since p-values are highly problematic and almost meaningless (and so are confidence intervals when used to see if zero is contained in them) why does it matter what threshold is used for them? Why is one interested in a null hypothesis in this situation? Do you really think it’s possible for the treatment to have zero effect to more than 10 decimal places?

Expand full comment