An article by Daryl Bem forthcoming (in a few months) in the Journal of Personality and Social Psychology, claims to have scientifically proven the existence of precognition. Many websites and newspapers have spoken, and the article is already criticized the methodological point of view by James Alcock in this article.
Other researchers have criticized the way Bem uses statistics, and advocate a Bayesian approach in the humanities. Their article should appear in the same journal.
I would add just a little note of importance, which greatly join those of Wagenmakers et al. In Experiment 1, Bem asked subjects to guess which side is an image that is sometimes erotic. In reality, there is no image before the subjects choose. Since the latter are in more than half of the erotic image, Bem concluded that precognition exists.
This result is valid at risk of 1% with one-tailed test. Surprising? No!
Bem because A does not test, but the equivalent of 16: First, it tests four sets of images to the same conclusion (the existence of psi). Each series is tested using 2 test (a test of Student, then a binomial test: we really wonder what the Student test doing here!). In addition, while an error rate well below 50% also show the existence of psi (one can imagine that people are reluctant to choose in an experiment erotic image, and avoids, for example ), Bem has chosen unilateral tests.
In the end, everything happens as if Bem was tested 16 times the same event ... he must correct the error for multiple testing, which led to accept the overall conclusion to 5% if one test is valid ... About 0.3%, meaning that none of the tests does Bem.
A re-analysis data Bem thus leads to the conclusion that they are in no way evidence of phenomena of precognition. The same phenomenon occurs on 9 (duffer!) Experiments reported in the article.
Besides if Bem was right, we should now understand why this wonderful effect that allows us to destroy the casino does not operate in such establishments.
No comments:
Post a Comment