Dear SPM-experts,
I have a question concerning DCM for induced responses. As far as I understood, the choice of the number of frequency modes is somewhat arbitrary, although the number should correspond to the number of frequency bands of interest. In the original paper of DCM for induced responses by Chen et al. it is recommended to use at least as many modes as needed to explain 90 % of the data’s variance.
My question is if using even more modes i.e. explaining up to 99 % of the data’s variance could be problematic for the interpretation of our subsequent model selection. Do the differences of different models only result from the 1% variance left unexplained?
Any help would be much appreciated.
Thank you and best regards,
Immo
|