A whopping 72 scientists have collaborated on a paper in which they advocate changing the default statistical significance threshold from 0.05 to 0.005. These luminaries of knowledge range from sociologists to mathematicians, from political scientists to economists, from biologists to statisticians, and they come from prestigious universities like Princeton, Harvard, Duke, the University of Milan, Ludwig-Maximilians-University Munich, and the Université Paris Dauphine to name but a few.
The group is concerned with reproducibility and credibility within scientific research and think that, in addition to publication bias and P-hacking, P < 0.05 is too low a bar for statistically significant results.
They recommend downgrading P-values of 0.05 from “significant” to “suggestive,” a change that Kelly Servick, writing for Science, breaks down for us casual statistical users.
The elite group of 72 does stress that results within their “suggestive” range are still scientifically useful and should be published in major journals, but will they get their way and change what we consider “significant”? Given the group’s collective resume, we think they’ve got a good shot, so we’ll keep you posted.