@maxmcd: a quick intro to publication bias: http://en.wikipedia.org/wiki/Publication_bias http://en.wikipedia.org/wiki/Publication_bias Imagine you're testing

@maxmcd: a quick intro to publication bias: http://en.wikipedia.org/wiki/Publication_bias http://en.wikipedia.org/wiki/Publication_bias Imagine you're testing the efficacy of a new drug, which has a close-to-neutral or neutral effect (to keep things simple), and there are n labs testing it. You might expect there to be a normal distribution of effect around a mean of zero (whatever quantity we're measuring). However, the groups who didn't find any effect decided to not publish their null results for any minor technicality or due to some pressure like (e.g. are you sure you don't want to repeat this experiment?) Left as an exercise: if you observe only at what got published, what do you expect to see? That being said, there's obviously more than just your garden variety publication bias and conflict of interest at play here.