该来源正文已进入翻译队列，中文正文生成前先展示摘要和原始出处入口。
Nature Mental Health, Published online: 24 July 2026; doi:10.1038/s44220-026-00686-6 Predicting symptom change is a key goal of machine learning in mental health. However, models can seem more accurate than they are due to regression to the mean, a common statistical effect often overlooked in machine learning. Here we outline its implications and a simple framework for separating genuine prediction from statistical artefact.
Regression to the mean inflates accuracy in machine learning prediction of symptom change
Nature Mental Health, Published online: 24 July 2026; doi:10.1038/s44220-026-00686-6 Predicting symptom change is a key goal of machine learning in mental health. However, models can seem more accurate than they are due to regression to the mean, a common statistical effect often overlooked in machine learning. Here we outline its implications and a simple framework for separating genuine prediction from statistical artefact.