VALISblog 


Thursday, September 01, 2005
Problems with scientific papers
An interesting article on possible problems with the accuracy of scientific journal articles. New Scientist reports a study which uses statistical analysis to show that there is less than a 50% chance that the results of any given scientific paper are true. While I don't have the scientific knowledge to argue this one, I would have thought that this is why we place an importance on replicability, and on metaanalysis. The probability that many papers which give the same result are all wrong is vanishingly small. The original article, by John Ionaddis, is in Public Library of Science: Medicine.  