Vankelecom Lara, Schacht Ole, Laroy Nathan, Loeys Tom, Moerkerke Beatrijs
Department of Data-Analysis, Ghent University, Belgium.
Department of Experimental Psychology, Ghent University, Belgium.
Psychol Belg. 2025 Jan 9;65(1):17-37. doi: 10.5334/pb.1318. eCollection 2025.
Performing hypothesis tests with adequate statistical power is indispensable for psychological research. In response to several large-scale replication projects following the replication crisis, concerns about the root causes of this crisis - such as questionable research practices (QRPs) - have grown. While initial efforts primarily addressed the inflation of the type I error rate of research due to QRPs, recent attention has shifted to the adverse consequences of low statistical power. In this paper we first argue how underpowered studies, in combination with publication bias, contribute to a literature rife with false positive results and overestimated effect sizes. We then examine whether the prevalence of power analyses in psychological research has effectively increased over time in response to the increased awareness regarding these phenomena. To address this, we conducted a systematic review of 903 published empirical articles across four APA-disciplines, comparing 453 papers published in 2015-2016, with 450 papers from 2020-2021. Although the prevalence of power analysis across different domains in psychology has increased over time (from 9.5% to 30%), it remains insufficient overall. We conclude by discussing the implications of these findings and elaborating on some alternative methods to power analysis that can help ensure sufficient statistical power.
进行具有足够统计效力的假设检验对心理学研究来说必不可少。针对复制危机之后的几个大规模复制项目，人们对这场危机的根源（如可疑的研究行为）的担忧与日俱增。虽然最初的努力主要解决了因可疑研究行为导致的研究I类错误率膨胀问题，但最近的关注点已转向统计效力低下的不良后果。在本文中，我们首先论证了效力不足的研究与发表偏倚相结合，是如何导致充斥着假阳性结果和高估效应大小的文献的。然后，我们考察了随着时间推移，心理学研究中功效分析的普及率是否因对这些现象的认识提高而有效增加。为解决这一问题，我们对美国心理学会四个学科领域的903篇已发表实证文章进行了系统综述，比较了2015 - 2016年发表的453篇论文和2020 - 2021年发表的450篇论文。尽管心理学不同领域功效分析的普及率随时间有所增加（从9.5%增至30%），但总体上仍显不足。我们通过讨论这些发现的意义并详细阐述一些功效分析的替代方法来结束本文，这些方法有助于确保足够的统计效力。
