Kurganskiĭ A V
Zh Vyssh Nerv Deiat Im I P Pavlova. 2010 Nov-Dec;60(6):740-59.
This review focuses on some practical issues of using vector autoregressive model (VAR) for multichannel EEG analysis. Those issues include: EEG preprocessing, checking if the necessary conditions of VAR model applicability are met, optimal order selection, and assessment of the validity of fitted VAR model. Both non-directed (ordinary coherence and imaginary part of the complex-valued coherency) and directed (directed coherence, directed transfer function and partial directed coherence) measures of the strength of inter-channel coupling are discussed. These measures are analyzed with respect to their properties (scale invariance) and known problems in using them (spurious interactions, volume conduction).
本综述聚焦于使用向量自回归模型(VAR)进行多通道脑电图(EEG)分析的一些实际问题。这些问题包括:EEG预处理、检查VAR模型适用性的必要条件是否满足、最优阶数选择以及对拟合VAR模型有效性的评估。文中讨论了通道间耦合强度的非定向(普通相干性和复值相干性的虚部)和定向(定向相干性、定向传递函数和部分定向相干性)测量方法。针对这些测量方法的特性(尺度不变性)及其使用中存在的已知问题(伪相互作用、容积传导)进行了分析。