Department of Clinical Child and Adolescent Psychology, Faculty of Psychology, University of Bergen, Bergen, Norway.
Department of Child and Adolescent Mental Health, Sørlandet sykehus, Kristiansand, Agder, 4615, Norway.
F1000Res. 2022 Sep 23;11:171. doi: 10.12688/f1000research.76842.2. eCollection 2022.
Statistical methods are a cornerstone of research in clinical psychology and are used in clinical trials and reviews to determine the best available evidence. The most widespread statistical framework, frequentist statistics, is often misunderstood and misused. Even when properly applied, this framework can lead to erroneous conclusions and unnecessarily prolonged trials. The implications for clinical psychology are difficulties in interpreting best available evidence and unnecessarily costly and burdensome research. An alternative framework, Bayesian statistics, is proposed as a solution to several issues with current practice. Statistical tests of primary outcome measures were extracted from 272 studies, which were cited in 11 recent reviews in the Evidence-based updates series in the Journal of Clinical Child and Adolescent Psychology. The extracted tests were examined regarding relevant features and re-analyzed using Bayes Factors. When statistical tests were significant, the majority (98%) of re-analyzed tests agreed with such claims. When statistical tests were nonsignificant almost half (43%) of re-analyzed tests disagreed with such claims. Equally important for clinical research, an average of 13% fewer participants per study would have been required if the studies had used Bayes Factors. Bayes Factors offer benefits for research in clinical psychology through intuitive interpretations, and less costly trials.
统计学方法是临床心理学研究的基石，在临床试验和综述中用于确定最佳现有证据。最广泛使用的统计学框架，即频率统计学，经常被误解和滥用。即使正确应用，该框架也可能导致错误的结论和不必要的延长试验。对临床心理学的影响是难以解释最佳现有证据，以及不必要的昂贵和繁琐的研究。贝叶斯统计学作为当前实践中的几个问题的解决方案被提出。主要结局测量的统计检验从在《临床儿童与青少年心理学杂志》的循证更新系列中最近 11 篇综述中引用的 272 项研究中提取出来。对提取的检验进行了相关特征的检查，并使用贝叶斯因子进行了重新分析。当统计检验显著时，重新分析的检验中有 98%（大多数）与此类说法一致。当统计检验不显著时，重新分析的检验中有 43%（几乎一半）与此类说法不一致。同样重要的是，对于临床研究，如果研究使用贝叶斯因子，每个研究需要的参与者平均减少 13%。贝叶斯因子通过直观的解释和成本更低的试验，为临床心理学研究提供了好处。
