Improving the Transparency and Robustness of Research

One of my first editorials as Editor-in-Chief offered guidelines on statistical reporting.1 This offered specific recommendations, including avoiding dichotomizing results into “significant” and “nonsignificant” and reporting effect size and confidence intervals. It was in part motivated by a desire to avoid what Gigerenzer2 has described as the use of “statistical rituals”—conducting statistical tests and generating p values without a clear sense of why. Since then, there has been considerable interest in initiatives intended to improve not only the quality of statistical reporting but the quality of how scientific studies are reported (and indeed conducted) more generally.