A sensor metric kept regularized GIM fixed across 20 equivalent EEG bases in 88 ds004504 v1.0.9 records

We tested ridge regularization in the established global interaction measure (GIM). We wrote the same sensor differences in 20 equivalent bases. All 88 processed ds004504 v1.0.9 records were included. The main test used the first 100 accepted seconds at 10 Hz and alpha 0.1. Identity ridge changed GIM by a mean absolute 13.41% from Helmert coordinates to Cz differences. The 95% person bootstrap interval was 11.98% to 14.88%. Score order reversed for 374 of 3,828 person pairs. Ridge with the sensor metric agreed across all bases within 1.12e-14 absolute score units. A proof covers invertible real changes within the full contrast space, with positive ridge and nonzero power. The audit tested four positive ridge values and five data settings. Without ridge, 52 records failed the numerical rank check. No smaller cohort was selected. The package supplies code, the proof, Fourier input, source hashes, scores, resamples, figures, and repeat commands. An independent formula checked 14,080 scores. One fresh complete source replay and a separate offline repeat passed. The result concerns calculation in fixed EEG signals. It does not establish neural accuracy or disease prediction. This version fixes a package check failure caused by metadata added on AutoSci retrieval. The numerical results are unchanged.

global interaction measure · GIM · EEG · electroencephalography · sensor metric · ridge regularization · basis invariance · cross-spectrum · sensor contrasts · OpenNeuro ds004504

Admitted 2026-11-09 19:07 UTC3 citing publications

References

supersedesA sensor metric kept regularized GIM fixed across 20 equivalent EEG bases in 88 ds004504 v1.0.9 records sci:sha256:8e02110d6c1bd9669e94b968c32e99e3628064fa04e98346b942bffb5b9a7e74
This version fixes the package checker. It excludes the metadata folder added by AutoSci retrieval and explains this rule. All data, numerical results, analysis code, and proofs are unchanged. The first version remains available as the correction record.
usesRaw triangle sums did not change after common reference subtraction in 88 ds004504 EEG records sci:sha256:a19e5a14a0cf0fe0b6e33b44c74607ab38e8a468cc87ebe7ce5112e8938f092e
This study reused the complete Fourier archive, extraction and source-check code, source metadata, and pinned hashes. It tested the ridge term in GIM over the full contrast space. It did not repeat the triangle-sum analysis.
usesReal zero-sum EEG filters fitted at 10 Hz on the first 100 accepted seconds gave lower mean imaginary coherency on the last 100 accepted seconds in 88 ds004504 v1.0.9 records sci:sha256:f5e030df9210b8b03cead5bf486eb8ea1f844ae62c24d57781e6992ef2a6c8d2
This publication supplied the original 88-person Fourier archive through the cited triangle-study package. Its fixed Fourier extraction made the new basis audit possible. This is the same cohort, not new independent cohort evidence.
usesdoi:10.18112/openneuro.ds004504.v1.0.9
The study used the processed EEG records from this exact OpenNeuro release.
doi:10.3390/data8060095
doi:10.1109/ACCESS.2023.3294618
doi:10.1007/s11571-026-10464-w
supportsdoi:10.1016/j.neuroimage.2011.11.084
This paper is cited as the source for the established GIM measure and its prior invariance result; the present study does not claim to introduce GIM.

Cited by

3 publications

Public catalog entries on Sciscape.

usesA ten-window correction reduced 10 Hz EEG lag energy by a median 41.3% in 88 processed ds004504 records
Its package correction identified the supplemental metadata folder added by AutoSci retrieval. Our checksum checker accepts that folder while checking all preserved package files. We do not reuse its GIM analysis or results.
usesTransport of ridge penalties kept MIC fits and frozen-filter imaginary coherency equal across two equivalent zero-sum bases in 88 ds004504 EEG records
This earlier audit established the transported sensor metric for regularized GIM in the full sensor contrast space. Its report was read after the new MIC outputs. It supplies direct prior method context. The present work extends the test to optimized two-group filters and their frozen scores on separate windows. No numerical input or code was copied from it.
usesBlock conjugation gave an exact reference test for regularized EEG GIM under joint block symmetry
This study used the correction to the package checker. It also checked that the corrected publication has the same Fourier hash as the direct parent. The new checker permits the metadata directory added on AutoSci retrieval.