mne_denoise.ssa.ssa_w_correlation#

mne_denoise.ssa.ssa_w_correlation(components: ndarray, window_length: int) → ndarray[source]#

Compute weighted correlations between SSA reconstructions.

Parameters:

  • components (array-like, shape (n_components, n_times)) – Reconstructed time series, typically returned by ssa_decompose().

  • window_length (int) – Embedding dimension used to produce components.

Returns:

correlation – Symmetric weighted-correlation matrix. Zero-energy components have a zero row and column.

Return type:

ndarray, shape (n_components, n_components)

Raises:

  • TypeError – If window_length is not an integer.

  • ValueError – If the component array or embedding dimension is invalid.

Notes

The weights equal the anti-diagonal multiplicities of the trajectory matrix. Magnitudes near zero indicate stronger separability; magnitudes near one indicate that two reconstructed components are strongly mixed. W-correlation is a diagnostic, not an artifact-selection rule [1].

References

[1]

Golyandina, N., & Zhigljavsky, A. (2013). Singular Spectrum Analysis for Time Series. Springer. https://doi.org/10.1007/978-3-642-34913-3