Multivariate dynamical systems models for estimating causal interactions in fMRI

► MDS uses a vector autoregressive state-space model incorporating both intrinsic and modulatory causal interactions. Intrinsic interactions reflect causal influences independent of external stimuli and task conditions, while modulatory interactions reflect context dependent influences. ► Causal interactions in MDS are modeled at the level of latent signals, rather than at the level of the observed BOLD-fMRI signals. ► We develop and compare two MDS estimation procedures using maximum likelihood estimation (MLE) and variational Bayesian (VB) approaches. ► Using extensive computer simulations, we demonstrate that, compared to Granger causal models, MDS exhibits superior performance. Among MDS methods, VB showed a superior performance at low SNRs and shorter time series.