Some people seem to think that the publication of the Witzum, Rips, and Rosenberg article in Statistical Science served as a stamp of scientific approval on the work. This is a great exaggeration. Statistical Science publishes a wide variety of papers of general interest to statisticians. Although the referees thought carefully about possible sources of error in the work, no one tried to reanalyze the data carefully and independently to try to uncover the presumed flaw in the logic.
The very few public statements I have made seem to have been misinterpreted to lend support to the notion that there may be some scientific basis for the findings of Witzum, Rips, and Rosenberg. My personal belief is quite the opposite: the authors' work did not go far enough to make me seriously think, even for a moment, that their results were anything other than coincidental, and likely due to a subtle flaw in their methodology. As I said in the preface to that issue of the journal, the paper was offered to our readers as a challenging puzzle. We published it in the hope that someone would step forward and do the careful analysis required to solve the puzzle, and that the discipline of statistics would be advanced through the identification of the logical errors in this kind of pattern recognition.
For additional information on this subject I recommend
Brendan McKay's website and also a very nice
article for a lay audience
by Maya Bar-Hillel and Avishai Margalit.
|Robert E. Kass|