This is a preprint version submitted to review for publication on the
IEEE Transactions on Biomedical Engineering (TBME).
In this study we investigated the estimation method of cross-bicoherence
to infer quadratic phase coupling (QPC) in interaction networks based on
empirical Bayes estimation. Nonlinear interaction, ubiquitous in a wide
range of biomedical fields, often makes it challenging to apply
conventional parametric approaches. While cross-bicoherence has been
recognized as a key statistic of QPC, prior studies have not elucidated
how reliable the estimation can be for realistic time series. This work
demonstrates that cross-bicoherence estimated by the proposed method
effectively detects QPC, with illustrative simulation studies of both
typical QPC and neural mass models of cortical columns.