Thursday, September 29, 2011

Statistical Significance and the Publication Process

In theory if you have an original presentation of data, or an original collective argument on a given scientific topic then, in order to get your ideas "out there" and heard among other scientists you submit your work to the peer review process. If the peer reviewers agree that the work holds merit then the work will be published. Unfortunately this publishing paradigm is an idealization - with journals commonly making subjective decisions as to what should be sent out for peer review.

I wrote a paper highlighting some of the common problems I see in the way standard scientists use statistics such as equating a statistically significant finding with a qualitatively significant one. I have not been able to find a journal willing to send the accompanying paper out for peer review, so instead of getting frustrated I've decided to simply post the article on the Cornell University Library (Arxiv). It doesn't get peer reviewed but at least scientists might read the article this way.

Just because an article should be read by many scientists out there, doesn't mean that there are any reputable publishing organizations willing to submit the content to the peer review process. An educational piece is not likely to generate citations and thus would be liable to adversely affect a journal's impact factor (a measure of how often a journal is cited and critical to a journal's perceived prestige within the scientific community).

Here's the link to the article's page at Cornell, and here's a link to the publicly hosted pdf file. If you have any comments feel free to let me know.

Jacob Levman, PhD