For arbitrary reasons. Today, given everything we know about easy frequentist statistics traps and caveats to fall into, p-hacking, “researcher degrees of freedom”, the replication crisis, etc. I’d honestly give the advice of ignoring anything above 0.001, and being suspicious of anything above 0.0001. And if you’re in a position to write a paper, pre-register, and use something more reasonable like Bayesian statistics.
Latest Answers