Gregory Francis on Publication Bias.
Neuroskeptic links to and comments on a new paper by Gregory Francis, and the issue of publication bias. Neuroskeptics post is, as always, worth reading. It is also worth reading Francis paper (NS links to the PubMed abstract, so it takes a few more clicks for the fortunate people inside the academic walls to get to the paper). Francis goes through first Bem’s now infamous paper, and then Schoolers talk of the decline effec. Both had some explanations for their results that would require some radical modification about how we believe the universe works. Francis shows why that is unlikely the case.
Francis is actually not that much for a pre-publication registry of planned research. Considering how often we really don’t know what we are doing anyway (he calls it exploratory. And, of course it is.) Or, more that when you are doing basic exploration, requiring that much work up front on things that can go wrong a thousand ways for no clear reason is maybe not the best use of our time. He suggests that we all learn Bayesian analysis instead, as he claims this can reliably demonstrate null effects also. I take his word for it. Time to crack open my Krushke textbook.