The standards of medical statistics

Bob Phair rphair at ix.netcom.com
Sun Mar 16 23:24:25 EST 1997


Everyone who has been an active scientist for more than a few years is
aware of many "errors" in the literature of his or her field. These
errors cover every aspect of experimental design, experimental methods,
data analysis and interpretation, not just statistical analysis.
Experience suggests it's impossible to assure that a paper is correct in
all particulars; we all know that two or three reviewers is an
insufficient number to evaluate every aspect of the paper. Furthermore,
no increase in the number of reviewers would be sufficient. Science is a
human endeavor and no small group of reviewers is blessed with enough
knowledge and insight to always identify the truth. Indeed, authors
often have excellent answers to what private reviewers thought were
devastating questions.

In some disciplines publication to the web is becoming the standard. It
seems to me that allowing other scientists to append their comments
(perhaps as hypertext links) to the web-published manuscript, with
digital signatures that cannot be forged, would go a long way toward
strengthening a desperately weak review process. At the same time, the
scientific community would benefit from a much broader review than is
now possible. Authors and others could, of course, respond to any such
critique, and the resulting exchange would greatly increase the
publication's value.

This approach could lead to a scientific literature that is more
approachable, more human, and more effective than a paper collection of
articles that have the imprimatur of just two reviewers and an editor.

--
Robert D. Phair, Ph.D.  rphair at ix.netcom.com
BioInformatics Services  http://www.webcom.com/rphair
Partnering and Outsourcing for Computational Biology





More information about the Comp-bio mailing list