mne_denoise.ssa.ssa_w_correlation

mne_denoise.ssa.ssa_w_correlation(components: ndarray, window_length: int) → ndarray

Compute weighted correlations between SSA reconstructions.

Parameters:

componentsarray-like, shape (n_components, n_times)

Reconstructed components.

window_lengthint

Embedding dimension used to obtain components.

Returns:

correlationndarray, shape (n_components, n_components)

Symmetric weighted-correlation matrix; zero-energy components have zero rows and columns.

Notes

The weights are the anti-diagonal multiplicities of the trajectory matrix. This is a separability diagnostic, not an artifact-selection rule. [1].

References

[1]

Nina Golyandina and Anatoly Zhigljavsky. Singular Spectrum Analysis for Time Series. SpringerBriefs in Statistics. Springer Berlin Heidelberg, 1 edition, 2013. ISBN 978-3-642-34913-3. doi:10.1007/978-3-642-34913-3.