When I was a young scientist (quite some time ago) there was
a joke that seemed to be circulating about how our older established colleagues
conducted science. This was a somewhat cynical exercise motivated, at least in
part, by professional jealousy. The joke went on to say that one could
establish a fact by writing two papers. In the first paper the author
speculates that something might be true. In the second paper the author says
that the previously speculated thing is true, and references the paper
containing the original speculation. In fact I have rarely seen this actually
done. But as I write blog I have an example sitting in front of me on my desk. It is especially intriguing that this paper was written by an individual who
maintains that “most published research findings are false”.
The paper in question was published just last month with the
rather presumptuous title: “How to make more published research true”. This, of course, is a statement predicated on
the presumption that much published research is false. Indeed the author says
in the first paragraph, referring to scientific research, that “Many new
proposed associations and/or effects are false or grossly exaggerated” and
refers to two previously published papers both single author papers by him.
So I should be able to go to these papers and see what
evidence that Dr. Ioannidis found which allows him to make such bold
conclusions. In the 2005 paper, however, I found no data. In fact it presents a
mathematical model using Bayesian statistics to simulate research outcomes and
finds that most studies using conventional Gaussian statistics and a p-value of
less than 0.05 are not adequate for identifying statistical significance.
The views, opinions and positions expressed by these authors and blogs are theirs and do not necessarily represent that of the Bioethics Research Library and Kennedy Institute of Ethics or Georgetown University.