Key points are not available for this paper at this time.
Type I error is a risk undertaken whenever significance tests are conducted, and the chances of committing a Type I error increase as the number of significance tests increases. But adjusting the alpha level because of the number of tests conducted in a given study has no principled basis, commits one to absurd beliefs and practices, and reduces statistical power. The practice of requiring or employing such adjustments should be abandoned.
Daniel J. O’Keefe (Tue,) studied this question.