Computational NeuroscienceAssessment of cross-frequency coupling with confidence using generalized linear models

•We use generalized linear models to assess cross-frequency coupling.•The method allows direct computation of confidence in the resulting statistic.•The method accurately detects biphasic cross-frequency coupling.•The resulting statistic is easily interpretable.

BackgroundBrain voltage activity displays distinct neuronal rhythms spanning a wide frequency range. How rhythms of different frequency interact - and the function of these interactions - remains an active area of research. Many methods have been proposed to assess the interactions between different frequency rhythms, in particular measures that characterize the relationship between the phase of a low frequency rhythm and the amplitude envelope of a high frequency rhythm. However, an optimal analysis method to assess this cross-frequency coupling (CFC) does not yet exist.New methodHere we describe a new procedure to assess CFC that utilizes the generalized linear modeling (GLM) framework.ResultsWe illustrate the utility of this procedure in three synthetic examples. The proposed GLM-CFC procedure allows a rapid and principled assessment of CFC with confidence bounds, scales with the intensity of the CFC, and accurately detects biphasic coupling.Comparison with existing methodsCompared to existing methods, the proposed GLM-CFC procedure is easily interpretable, possesses confidence intervals that are easy and efficient to compute, and accurately detects biphasic coupling.ConclusionsThe GLM-CFC statistic provides a method for accurate and statistically rigorous assessment of CFC.