Table 1.

Overview of methods.

DOI: http://dx.doi.org/10.7554/eLife.21792.002

	Description	Analysis goal	Key assumption
Method 1	S defined by peri-LF-peak; R defined by all data	Identify a single phase-amplitude coupled network	One HF network with power proportional to LF phase
Method 2	S defined by peri-LF-peak; R defined by peri-LF-trough	Identify two networks that alternate according to LF phase.	Two different HF networks that have power peaks at different LF phases
Method 3	LF activity bias-filters sphered data	Use (possibly nonstationary) LF waveform shape to identify a HF component.	Well-defined LF waveform
Method 4	Delay-embedded matrix, S and R defined as in Methods 1 or 2	Empirically determine a CFC-related spatiotemporal filter	Appropriate delay-embedded order
Method 5	Similar to Method 4 but data are taken peri-action potential	Empirically determine a spatiotemporal LFP filter surrounding action potentials	Sufficient delay-embedding order; one peri-spike network

Notes. LF = low frequency; HF = high frequency. All methods make the assumption that the spatiotemporal characteristics of the HF activity are stable over repeated time windows from which covariance matrices are computed.