Department of Cognitive Science, University of California-San Diego, La Jolla, CA 92093-0515, USA.
Psychophysiology. 2011 Dec;48(12):1711-25. doi: 10.1111/j.1469-8986.2011.01273.x. Epub 2011 Sep 6.
Event-related potentials (ERPs) and magnetic fields (ERFs) are typically analyzed via ANOVAs on mean activity in a priori windows. Advances in computing power and statistics have produced an alternative, mass univariate analyses consisting of thousands of statistical tests and powerful corrections for multiple comparisons. Such analyses are most useful when one has little a priori knowledge of effect locations or latencies, and for delineating effect boundaries. Mass univariate analyses complement and, at times, obviate traditional analyses. Here we review this approach as applied to ERP/ERF data and four methods for multiple comparison correction: strong control of the familywise error rate (FWER) via permutation tests, weak control of FWER via cluster-based permutation tests, false discovery rate control, and control of the generalized FWER. We end with recommendations for their use and introduce free MATLAB software for their implementation.
事件相关电位（ERPs）和磁场（ERFs）通常通过在先验窗口中对平均活动进行方差分析来进行分析。计算能力和统计学的进步产生了一种替代方法，即由数千个统计检验和强大的多重比较校正组成的大规模单变量分析。当对效应位置或潜伏期几乎没有先验知识，并且需要描绘效应边界时，这种分析最有用。大规模单变量分析补充并有时替代传统分析。在这里，我们回顾了这种方法在 ERP/ERF 数据中的应用，以及四种多重比较校正方法：通过置换检验对总体错误率（FWER）进行强控制、通过基于聚类的置换检验对 FWER 进行弱控制、错误发现率控制和广义 FWER 控制。最后，我们对它们的使用提出了建议，并介绍了用于实现它们的免费 MATLAB 软件。
