'Statistical Significance' Is Overused And Often Misleading : Shots - Health News Scientists and statisticians are putting forth a bold idea: ban the very concept of "statistical significance." A bit more humility would be in order to account for the ambiguity in the world.

Statisticians' Call To Arms: Reject Significance And Embrace Uncertainty!

  • Download
  • <iframe src="https://www.npr.org/player/embed/705191851/706295436" width="100%" height="290" frameborder="0" scrolling="no" title="NPR embedded audio player">
  • Transcript


Eggs are bad for you, according to a study out earlier this month. But wait. Before that, eggs were good for you. And before that, they were bad. This is not because the truth is changing all the time. One key reason - scientists have a hard time coping with uncertainty. Don't we all? NPR's Richard Harris reports on a new effort to break science out of that confusing rut.

RICHARD HARRIS, BYLINE: One of the people who's asking scientists to reconsider how they treat uncertainty is Nicole Lazar, a professor of statistics at the University of Georgia. How does she react to the latest study of alcohol, coffee, eggs or whatever?

NICOLE LAZAR: Whenever I see anything in the paper or my husband tells me something about, oh, a new study shows - I'm just like, whatever. You know, I don't even pay attention to it anymore.

HARRIS: That's not because she's fatalistic. It's because science - the way it's practiced today - actually encourages scientists to boil everything down to a true/false question.

LAZAR: The real world is much more uncertain than that.

HARRIS: Of course, scientists are drawn toward specialized language. So you hear their rendition of true all the time on NPR and elsewhere.


UNIDENTIFIED PERSON: And that difference was statistically significant.

MICHAEL POLLAN: And they found that 80 percent of the people in the trial had statistically significant reductions.

RACHEL HINNENKAMP: That's a statistically significant increase.

HARRIS: Statistically significant is equated with true or real. Though, that's really not the case. Lazar says it's certainly convenient to have an easy shortcut that, seemingly, helps distinguish strong results from forgettable ones.

LAZAR: Having that bright-line cutoff makes everything seem much more certain than it really is.

HARRIS: It actually distorts the truth. Sometimes, scientists actually play games with this bright-line, massaging their data to make sure it lands just barely on the desirable side. Other times, people ignore findings that actually might deserve a second look. So Lazar is among a group of more than 800 scientists who are saying it's time to abolish the badly abused concept of statistical significance.


RON WASSERSTEIN: It's time to stop using that phrase. It's really gotten stretched all out of proportion.

HARRIS: Ron Wasserstein is executive director of the American Statistical Association, and he's been arguing this for years. But it's deeply embedded in the world of science. Journals demands statistical significance. College deans count on it, so do grant reviewers. But there are dangers of continuing to use this intellectual shortcut.


WASSERSTEIN: Failure to make these changes are now really starting to have a sustained negative impact on the way science is conducted. And it's time to make the changes. It's time to move on.

HARRIS: Scientists bury perfectly good data because they aren't statistically significant, he says. And studies can easily end up with the wrong conclusions after being forced through this abused test. His association's journal, American Statistician, has just published 43 papers decrying the practice and discussing alternatives. Wasserstein says one thing scientists should do is embrace uncertainty rather than using statistics to sweep it under the rug.


WASSERSTEIN: Uncertainty is present - always. That's part of science. So rather than try to dance around that, we accept it.

HARRIS: Measure it and make better use of it. Wasserstein says this goes against human nature because we want answers not perpetual questions. And some statistics experts say we shouldn't ditch this flawed system until we know what replaces it will actually be an improvement. But Wasserstein says dropping the concept of statistical significance gives you a more honest way of looking at research, like the egg study, which is surely not the last word on a messy question about nutrition.

Richard Harris, NPR News.


Copyright © 2019 NPR. All rights reserved. Visit our website terms of use and permissions pages at www.npr.org for further information.

NPR transcripts are created on a rush deadline by an NPR contractor. This text may not be in its final form and may be updated or revised in the future. Accuracy and availability may vary. The authoritative record of NPR’s programming is the audio record.