Data quality and reliability metrics for event-related potentials (ERPs): The utility of subject-level reliability
Data quality and reliability metrics for event-related potentials (ERPs): The utility of subject-level reliability
复制标题
DOI：
10.1016/j.ijpsycho.2021.04.004
复制
发表时间：
2021-05-03
影响因子：
3
通讯作者：
Hajcak, Greg
中科院分区：
文献类型：
--
作者：
Clayson, Peter E.;Brush, C. J.;Hajcak, Greg
事件相关脑电位(ERPs)代表神经活动的直接测量，被用来理解认知、情感、感觉和运动过程。每一位ERP研究人员都会遇到一个障碍，那就是确定测量是否足够精确或心理测量学是否足够可靠，以达到预期的目的。在这本入门读物中，我们回顾了三种类型的度量指标：数据质量、小组级别的内部一致性和主题级别的内部一致性。数据质量评估表征了ERP分数的精确度，但没有提供关于分数是否足够精确来检验个体差异的固有信息。群体层面的内部一致性表征了人与人之间的差异与这些分数的精确度的比率，并为整个参与者提供了单一的内部一致性估计，这可能会掩盖某些个人的内部一致性较低的风险。受试者级别的内部一致性考虑了一个人的ERP分数相对于一个组的人与人之间的差异的精确度，并为每个人产生了估计。我们将每个度量应用于已发布的错误相关负面(ERN)和奖励积极(REWP)数据，并展示了不考虑数据质量和内部一致性如何破坏统计推断。最后，我们对如何使用这些估计来提高测量质量和方法透明度提出了一般性意见。主题级内部一致性计算在ERP可靠性分析(ERA)工具箱中实施。
Event-related brain potentials (ERPs) represent direct measures of neural activity that are leveraged to understand cognitive, affective, sensory, and motor processes. Every ERP researcher encounters the obstacle of determining whether measurements are precise or psychometrically reliable enough for an intended purpose. In this primer, we review three types of measurements metrics: data quality, group-level internal consistency, and subject-level internal consistency. Data quality estimates characterize the precision of ERP scores but provide no inherent information about whether scores are precise enough for examining individual differences. Group-level internal consistency characterizes the ratio of between-person differences to the precision of those scores, and provides a single internal consistency estimate for an entire group of participants that risks masking low internal consistency for some individuals. Subject-level internal consistency considers the precision of an ERP score for a person relative to between-person differences for a group, and an estimate is yielded for each individual. We apply each metric to published error-related negativity (ERN) and reward positivity (RewP) data and demonstrate how failing to consider data quality and internal consistency can undermine statistical inferences. We conclude with general comments on how these estimates may be used to improve measurement quality and methodological transparency. Subject-level internal consistency computation is implemented within the ERP Reliability Analysis (ERA) Toolbox.