However, MI-values from the MVL-MI-Canolty algorithm have been shown to partly reflect the power of fa oscillations, rather than their coupling (Canolty and Knight, 2010). Therefore, as an alternative to surrogate data, we applied a MVL-MI algorithm from Özkurt and Schnitzler (2011), which includes a normalization factor corresponding to the power of fa. Özkurt and Schnitzler (2011) suggest that their algorithm is more resilient to measurement noise, and is therefore highly relevant for MEG data, which has an inherently lower signal-to-noise ratio compared with invasive electrophysiological recordings (Goldenholz et al., 2009).
