Researchers continue to debate the role of the much-criticized p-value, Nature News reports.
Back in July, a group of psychology researchers called in a paper appearing at the PsyArXiv preprint server for the cut-off of deeming a finding 'statistically significant' — typically a p-value of 0.05 — to be lowered. The researchers argued that the higher p-value has allowed too many false positives to creep into the literature. Instead, they suggested that the social and biomedical sciences use a p-value of 0.005.
In a new response, also at PsyArXiv, another group of researchers argues instead that researchers should rather explain why they chose the p-value they did, rather than rely on a blanket value. "Instead of a narrower focus on p-value thresholds, we call for a broader mandate whereby all justifications of key choices in research design and statistical practice are pre-registered whenever possible, fully accessible, and transparently evaluated," the team writes.
Ronald Wasserstein, executive director of the American Statistical Association, tells Nature News that while he hasn't taken a side in the debate, he says "we shouldn't be surprised that there isn't a single magic number."