Is neuroscience facing up to statistical power?

TL;DR

This paper reviews the issue of low statistical power in neuroscience studies, highlighting that most recent research does not adequately address this problem despite its implications for reproducibility.

Contribution

It provides an analysis of sample size justifications in recent neuroscience publications, revealing a widespread lack of proper power considerations.

Findings

Only 1 of 15 papers claimed adequate power

Most studies relied on previous research without validation

Concerns about statistical power remain largely unaddressed in neuroscience

Abstract

It has been demonstrated that the statistical power of many neuroscience studies is very low, so that the results are unlikely to be robustly reproducible. How are neuroscientists and the journals in which they publish responding to this problem? Here I review the sample size justifications provided for all 15 papers published in one recent issue of the leading journal Nature Neuroscience. Of these, only one claimed it was adequately powered. The others mostly appealed to the sample sizes used in earlier studies, despite a lack of evidence that these earlier studies were adequately powered. Thus, concerns regarding statistical power in neuroscience have mostly not yet been addressed.