Critical Values Robust to P-hacking
This paper builds a model of hypothesis testing with p-hacking and gives critical values that correct the inflated type 1 error rate caused by p-hacking. As a rule of thumb, such robust critical values are classical critical values with one fifth of the significance level.