Positive reasons for publishing negative findings

Abstract

Scientific and medical authors tend to be biased toward submitting "statistically significant" findings for publication. Journals show a similar bias in publishing such "positive" studies. The large number of publications in medical research means that, in a field obsessed with controlling Type I error rates, we have journals with an abundance of Type I errors. Failing to publish studies that do not show a treatment or exposure effect creates a literature conspicuously absent of trials necessary for unbiased meta-analyses and systematic reviews. Furthermore, by shelving or rejecting studies with nonstatistically significant outcomes, authors and editors censor the most important contributors to medical research: our consenting volunteers.