Reminds me of the idea (Robin Hanson's, I think?) to add an extra layer of blindness to studies: during peer review, take the original data, and write a separate paper with the opposite conclusion. Randomize which reviewers get which version. Your original paper is then only accepted if they reject the inverted version.
