Login to Your Account

A Likely Story

Current ‘P’ Values Are Far Too Lax, Study Suggests

By Anette Breindl
Science Editor

Tuesday, November 12, 2013
To be of any use at all, scientific experiments need to be reproducible. And it has become increasingly clear over the past few years that far too many of them aren’t. There are a number of suggestions for how to fix that low reproducibility rate, including blinding preclinical experiments in the same way as clinical trials. But in the Nov. 11, 2013, issue of the Proceedings of the National Academy of Sciences, statistician Valen Johnson provided evidence of another problem for reproducibility: It may be too easy to label findings significant in the first place.

To continue reading subscribe now to Science News

Learn More about Science News

Already a subscriber? Sign In or Buy now to activate your subscription