IDTxl

IDTxl infers directed networks and their dynamics from multivariate time series using multivariate transfer entropy to detect nonlinear and lagged dependencies.

Key Features:

• Multivariate Transfer Entropy: Employs multivariate transfer entropy as a model-free measure to identify directed, lagged, and nonlinear interactions among time series.
• Hierarchical Statistical Testing: Implements hierarchical statistical tests that control the family-wise error rate and support parallelized computation.
• Greedy Selection Algorithms: Uses greedy algorithms to select minimal directed network models while avoiding redundant inferences and capturing synergistic effects among variables.
• Scalability and Efficiency: Validated on synthetic networks up to 100 nodes and optimized to handle datasets an order of magnitude larger in network size and sample size, suitable for EEG and magnetoencephalography experiments.
• Performance Validation: Demonstrated high precision, recall, and specificity (>98% on average) on synthetic time series of 10,000 samples across linear and nonlinear dynamics, with improved precision–recall trade-offs for longer time series.

Scientific Applications:

• Neuroscience and Neuroimaging: Inferring functional and effective connectivity in neuroimaging datasets such as EEG and magnetoencephalography to study complex neural interactions.

Methodology:

IDTxl applies multivariate transfer entropy together with greedy selection algorithms and hierarchical statistical testing (controlling the family-wise error rate) and enables parallelized computation.