Directed Transfer Function (Kaminski & Blinowska 1991) between every channel pair, derived from a fitted MVAR transfer function. More...
#include "../connectivity_global.h"#include "abstractmetric.h"#include <QSharedPointer>#include <Eigen/Core>

Go to the source code of this file.
Classes | |
| class | CONNECTIVITYLIB::DirectedTransferFunction |
| Directed Transfer Function estimator (Kaminski & Blinowska 1991); directional, includes indirect paths. More... | |
Namespaces | |
| namespace | CONNECTIVITYLIB |
| Functional connectivity metrics (coherence, PLV, cross-correlation, etc.). | |
Directed Transfer Function (Kaminski & Blinowska 1991) between every channel pair, derived from a fitted MVAR transfer function.
SPDX-License-Identifier: BSD-3-Clause Copyright (c) 2026 MNE-CPP Authors
The DTF normalises each row of the squared MVAR transfer matrix to a unit sum,
DTF_{ij}(f) = |H_{ij}(f)|^2 / sum_k |H_{ik}(f)|^2
which can be read as the fraction of the spectral power that channel i would have at frequency f that originates from channel j relative to all other channels in the model. Output is non-negative, bounded by 1, and asymmetric (DTF_{ij} != DTF_{ji} in general). Kaminski and Blinowska (Biological Cybernetics, 1991) introduced DTF for multi-channel EEG to track stimulus-driven propagation patterns across cortex, and it remains a popular reference for directed flow in multivariate spectral analysis.
Unlike PartialDirectedCoherence, DTF includes both direct and indirect causal pathways: an indirect interaction j -> k -> i shows up as a non-zero DTF_{ij}. The two metrics are therefore complementary and are usually reported alongside GrangerCausality, with all three computed from the same MvarModel fit.
Definition in file directed_transfer_function.h.