Dear Alex,
The modes in DCM for ERP represent channel mixtures based on SVD of the channel data (for more details, see Kiebel, David & Friston, Dynamic causal modelling of evoked responses in EEG/MEG with lead field parameterization, Neuroimage 2005).
The number of modes is user-dependent. In principle, these should capture the main variance components associated with your signal and exclude variance due to noise. One way of choosing the number of modes is to use as many modes as there are sources. This is because the number of sources limits the dimensionality of the subspace of sensor space that you can model (cf. Garrido et al., Dynamic causal modelling of evoked potentials: A reproducibility study, Neuroimage 2007).
To assess whether percentage of explained variance is reasonable, you can calculate R2 (coefficient of determination) based on the fields DCM.R and DCM.H of your inverted models; see spm_dcm_erp_results, lines 419-500.
Best,
Ryszard
|