Centre for Cognitive Neuroimaging (CCNi), Institute of Neuroscience and Psychology, College of Medical, Veterinary and Life Sciences, University of Glasgow, Glasgow, UK.
Neuroimage. 2011 Sep 15;58(2):620-9. doi: 10.1016/j.neuroimage.2011.06.052. Epub 2011 Jun 25.
A reliable measure is one we can trust in the long run. Thus, the reliability of measurements is as important as their validity. Here we investigated the reliability of brain electrical visual evoked responses to faces and noise textures. For the first time, we provide reliability measures for the full time course of event-related potentials (ERPs). Our analyses were also performed on a R(2)(t) metric that reflects results from single-trial analyses, therefore providing the first reliability analysis of ERP single-trial analyses. Results show that ERPs and R(2)(t) are highly reliable (cross-correlation ~0.9, lag ~4/6ms, intra-class correlation ~0.9) but also idiosyncratic: ERPs and R(2)(t) are highly reproducible within subjects, who differ reliably from each other and the grand average across subjects. Consequently, grand averages, although highly reliable, can be misleading because they might not reflect the actual brain dynamic of any subjects.
一个可靠的测量方法是我们可以长期信赖的方法。因此，测量的可靠性与有效性同样重要。在这里，我们研究了大脑对人脸和噪声纹理的电视觉诱发电反应的可靠性。我们首次为事件相关电位（ERP）的整个时程提供了可靠性度量。我们的分析还基于反映单试次分析结果的 R(2)(t)指标，因此提供了对 ERP 单试次分析的可靠性分析。结果表明，ERP 和 R(2)(t)具有高度可靠性（相关系数0.9，滞后4/6ms，内类相关系数~0.9），但也具有个体差异：ERP 和 R(2)(t)在个体内高度可重现，个体之间可靠地相互区分，并且与跨个体的总和平均值区分开来。因此，总和平均值虽然高度可靠，但可能会产生误导，因为它们可能无法反映任何个体的实际大脑动态。
