Hi Christian,
indeed, I do use the --desnorm option to normalize the time courses/design matrix to unit variance. Essentially, to compute the residuals, I then used the normalized (not the original ones) courses:
residuals = data_masked - Tmodes_norm * Spatial_modes
Shouldn't the removal of the variance scaling for the time courses finally be reflected in the intensity/scaling of the obtained spatial maps anyway? In order to fit the data with minimal error?
Cheers,
Tobias
Dipl.-Ing. Tobias Wissel, M.Sc. Embedded Systems
PhD Student
UNIVERSITY OF LUEBECK
INSTITUTE FOR ROBOTICS AND COGNITIVE SYSTEMS
Germany
|