Hierarchical control of false discovery rate for phase locking measures of EEG synchrony

Computing phase-locking values (PLVs) between EEG signals is becoming a popular measure for quantifying functional connectivity, because it affords a more detailed picture of the synchrony relationships between channels at different times and frequencies. However, the accompanying increase in data dimensionality incurs a serious multiple testing problem for determining PLV significance. Standard methods for controlling Type I error, which treat all hypotheses as belonging to a single family, can fail to detect any significant discoveries. Instead, we propose a novel application of a hierarchical FDR method, which subsumes multiple families, for detecting significant PLV effects. For simulations and experimental data, we show that the proposed hierarchical FDR method is most powerful. This method revealed significant synchrony effects in the expected regions at an acceptable error rate of 5%, where other methods, including standard FDR correction failed to reveal any significant effects.