Fig. 6.

An example of dynamic effective connectivity estimation. (A) Predicted (solid lines) and observed (dotted lines) auto-spectra (ASD) of eight nodes of the default mode network are shown. (B) For the connection from MTG to IFG, predicted CSD with a dynamic model (red line) was very similar to the observed CSD (dotted line) for temporal window (#3) but differed significantly from the averaged CSD, across windows, in a stationary model (black line). (C) and (D) present the dynamics of both observed and predicted CSDs of the connection from MTG to IFG, which was estimated at the first level. Color levels indicate log-transformed CSD powers. (E) presents a time course of effective connectivity estimates (posterior expectations as the red line and confidence intervals in grey) of the same connection from MTG to IFG, with the window index on the x-axis. (F) shows the time course (black line) of effective connectivity of the same connection (IFG ← MTG) but estimated at the second level with a full (six basis functions) design matrix of DCT⁶ (2nd Xfull). (F) reproduces the slow fluctuations (dotted lines) in the first level DCM estimators (in E), time courses of predicted connectivity with the first and second DCT basis function (2nd X1:2), and time courses of predicted connectivity with (up to) four DCT basis functions (2nd X1:4).