The American Statistical Association wants to get researchers to change their p-value habits, New Scientist writes.
Researchers often rely on getting a p-value below 5 percent to determine whether their results are statistically significant. As New Scientist notes, the assumption is that when a p-value falls below that threshold, the results are not likely to be a coincidence. But, it adds, that that's not quite the case, and that this reliance on p-values may be contributing to the issue of scientific studies not being reproducible.
The ASA also warned last year about the misuse and misinterpretation of p-values.
However, New Scientist says there's no agreement to what should be done. One team recently suggested that a more stringent threshold of 0.005 be adopted, though another group countered that researchers should instead say in their papers why they chose the cut-off value they did, rather than relying on a conventional value.
New Scientist advises that statisticians at their annual meeting to try to reach a solution. "Unless they do, researchers will simply stick with the flawed methods they've always used — and leave the rest of us struggling to tell the fake breakthroughs from the real ones," it adds.