Prevention of false positive findings in observational studies: registration will not work but replication might

Hypotheses are often generated iteratively reflecting developments in the scientific literature or earlier findings within the cohort itself. [...]a procedure would still be vulnerable to manipulation once researchers have data in hand, because they could still engage in 'data-fishing'. [...]many researchers may not be willing to take the risk that their hypotheses might be 'stolen' by others when made public. [...]a registration requirement of observational studies would not appear to be a simple solution, nor would it be fully effective in preventing the spread of false positive findings.