Fig. 2.

Second level models for dynamic effective connectivity analysis. For the whole time series of the default mode brain regions (ROI₁,$\cdot \cdot \cdot$, ROI₈), a spDCM was estimated under the assumption of the stationarity during the session (A). To characterize the dynamics of the intrinsic effective connectivity, we constructed four types of second level models; a simple average model, i.e., Bayesian average of DCMs for all windows (B), a DCT model with DCT basis function as regressors (C), PCA (D) and fPCA models (E) that were based on the initial estimates of intrinsic effective connectivity.