Sometimes R built-in statistical test functions such as `cor.test`

report a p-value of 0, which makes me wonder the smallest p-value that can precisely be returned by these functions. I think this should be smaller that 1e-21, because I've seen something around 3e-21 returned.

This thread might help address your question: https://stackoverflow.com/questions/35349388/precision-of-scientific-notation-in-r

1 Like

You must have a very special kind of question to be worrying about numbers like 3e-21.

There are sources of uncertainty in p-values that don't stem from computer floating-point arithmetic. See, for example, this paper.

1 Like

This topic was automatically closed 21 days after the last reply. New replies are no longer allowed.