Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

Compared with many things we believe we know, p<0.05 is still pretty weak.

For instance, I've seen quite a few medical correlations that were P<0.001.

I try now to avoid the mindset of "significant" vs "insignificant" as having a discrete black or white cutoff.



p<0.05 is a ridiculous high bar for speculative research, and a ridiculous low bar for practical research.

If I didn't know its rationale, it does look a lot of what a mixed committee would settle on trying to appease everyone. But yeah, it's not a good static value.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: