Is it time for a Journal of Reproducible Research? Experiments that are independently conducted in different labs that find the same outcome. This topic has been on my mind recently, and the question of reproducibility and fraud has also been in the news. Such a journal would protect against one-off results and science fraud.We talked about fraud in the training seminar last week, discussing two papers (one retracted by the journal and the other not) known to be fraudulent. We boiled fraud down to three types:
- statistical trickery,
- altering data labels, and
- complete fakery of the data, through RNGs and the like.
We decided there is currently no way to independently verify whether data is real or not (but we decided theoretical studies harder to fake). The social evolution people came to the depressing conclusion that the best strategy for being a successful scientist is to cheat, if you have a mental fortitude to withstand the guilt.
Fraud and one-offs will continue until there is a significant benefit to catching fraudsters, of which there is currently none. This could be accomplished if only results that are independently verified are cited–a “Science 2.0”, if you like, a revolution in the execution of science. (As long as paired-up research groups aren’t in collusion with each other…)