Assessing dynamic spectral causality by lagged adaptive directed transfer function and instantaneous effect factor.

Area of Science:

• Neuroscience
• Signal Processing
• Computational Biology

Background:

• Assessing dynamic spectral causality in physiological signals is crucial.
• Time-varying multivariate autoregressive (tvMVAR) models and spectral Granger causality are common frameworks.
• Distinguishing directed instantaneous causality is challenging with Gaussian residuals and missing causal ordering information.

Purpose of the Study:

• To propose a new algorithm for assessing time-varying causal ordering in tvMVAR models.
• To estimate the instantaneous effect factor (IEF) for tracking dynamic directed instantaneous connectivity.
• To develop a time-lagged adaptive directed transfer function (ADTF) for lagged causality assessment after removing instantaneous effects.

Main Methods:

• Developed a novel algorithm to estimate time-varying causal ordering under the assumption of consistent acyclic causal ordering.
• Estimated the instantaneous effect factor (IEF) to quantify instantaneous connectivity.
• Applied a time-lagged ADTF to assess lagged causality post-instantaneous effect removal.
• Validated through simulations and real visual evoked potential (VEP) data.

Main Results:

• The proposed algorithm accurately estimates causal ordering and IEF values in Gaussian residual conditions.
• The time-lagged ADTF approach shows improved accuracy in estimating dynamic interactions in complex nervous systems.
• The method demonstrated effectiveness in assessing time-variant spectral causality on real VEP data.

Conclusions:

• The new algorithm provides accurate causal ordering and IEF estimation for tvMVAR models, particularly with Gaussian residuals.
• The time-lagged ADTF enhances the analysis of dynamic interactions by accounting for instantaneous effects.
• The method is valuable for investigating mutual causality and brain activity, as shown in VEP experiments.