When performing intra-cortical connectivity analysis, why is it important to orthogonalize the data prior to computing magnitude square coherence between electrode pair?
I wouldn’t say it’s generally important. As far as I know, time-domain orthogonalization is used only in one (out of many) correlation analysis methods from Hipp et al. There the idea is to remove any shared activity that could be attributable to volume conduction.
If I am not planning on computing the surface laplacian, and would like to remove the task based volume conduction (Grefkes et al). Then time-domain orthogonalization might be a good idea? My objective is to create a connectivity matrix (electrode x electrode) using magnitude square coherence.
Sure, you could try it. I’ve actually never tried that method myself, so I cannot say anything from personal experience; just that it’s a published method.
Keep in mind that time-domain orthogonalization is used for time-domain correlation analyses (e.g., correlated power time series). Magnitude squared coherence is a frequency-domain analysis. Perhaps a good option is to use the imaginary part of the coherence. That eliminates volume conduction, and it’s more commonly used and has been around longer.