Unbiased estimator of squared Phase Lag Index between every channel pair. More...
#include "../connectivity_global.h"#include "abstractmetric.h"#include "../connectivitysettings.h"#include <QSharedPointer>#include <QMutex>#include <Eigen/Core>

Go to the source code of this file.
Classes | |
| class | CONNECTIVITYLIB::UnbiasedSquaredPhaseLagIndex |
| Unbiased squared Phase Lag Index estimator; removes the small-sample bias of PhaseLagIndex. More... | |
Namespaces | |
| namespace | CONNECTIVITYLIB |
| Functional connectivity metrics (coherence, PLV, cross-correlation, etc.). | |
Unbiased estimator of squared Phase Lag Index between every channel pair.
SPDX-License-Identifier: BSD-3-Clause Copyright (c) 2018-2026 MNE-CPP Authors
The plain Phase Lag Index (Stam et al. 2007) and its square are biased upward when the number of trials N is small: even uncorrelated signals produce a non-zero expected magnitude of the empirical sign average because each sign() contribution has variance 1. The unbiased squared PLI removes the diagonal contribution analytically,
USPLI_{xy}(f) = ( ( sum_n sign(Im(S_{xy}^{(n)}(f))) )^2 - N ) / ( N * (N - 1) )
which is an unbiased estimator of PLI^2 in the sense that E[USPLI] = PLI^2 for any N >= 2. Values outside the nominal [0, 1] range can occur for very small N or noise-dominated pairs; these indicate a true PLI indistinguishable from zero at the available sample size and are usually clipped to 0 for display purposes.
This estimator is the direct phase-only analogue of the squared debiased weighted PLI of Vinck et al. (2011) and is the right choice when amplitude weighting is undesirable but small-sample bias must be removed.
Definition in file unbiasedsquaredphaselagindex.h.