How not to do Science – Non-climate Science Edition

I really do not want to know. Part of the point of having two independent experiments is that they cross-check each other - independently. We do that most effectively when we are blind to the other experiment's data, right up to the last minute. In fact we even try to blind ourselves to our own data up to a point. As much as possible of the analysis should be optimised and decided in advance, before even looking at the data. This prevents even the possibility of subconscious bias entering the studies. If you are biased, the truth will probably still out in the end, but in the meantime your statistical estimations of confidence and significance will all be wrong.