I’m trying to apply a Gaussian taper and Graphical LASSO L1 Regularization to my windows for dFC. Based on looking through toolbox codes, demeaning seems to be necessary to avoid spurious correlations for tapering. On the other hand, demeaning and dividing by the std seems to be necessary for regularization. What should be the correct order in applying the taper, demeaning, and dividing by the std?
E.g. windowed timeseries > demean > divide by std > taper > regularization?
E.g. windowed timeseries > demean > taper > demean > divide by std > regularization?