Statistical Biases in Science Communication

Misuse and misinterpretation of statistics result in statistical biases that affect the quality, clarity, relevance, and implications of communicated scientific information. Statistical tools are often suboptimally used in scientific papers, even in the best journals. The vast majority of published results are statistically significant, and even nonsignificant results are often spun as being important. Inferences based on P-values generate additional misconceptions. It is also common to focus on metrics that are more prone to exaggerated interpretation. Most of these problems are possible to solve or at least improve on. The prevalence of statistical biases has been used in attacks designed to discredit science’s validity. However, the use of rigorous statistical methods and their careful interpretation can be one of the strongest distinguishing features of good science and a powerful tool to sustain science’s integrity.