Recalculated contact variances changed 1–80 Hz dimension scores for unchanged bipolar SEEG in 36 patients

We tested a contact variance metric with public SEEG from 36 patients. The source was OpenNeuro ds004100, version 1.1.3. The primary band was 1–80 Hz. Global common-average reference changes preserved the bipolar signals. Recalculating contact variances changed normalized generalized participation ratio. The mean absolute patient change was 0.01759. The 95% patient bootstrap interval was 0.01232–0.02370. The mean score changed from 0.18821 to 0.19965. A fixed metric preserved the score within 6.8e-15. A three-contact counterexample proves that this recalculated score is not determined by bipolar covariance alone. We also tested shaft averages, two secondary bands, and added common variance. Fresh checks rebuilt all 72 pinned public records with exact covariance agreement. The package includes code, 432 covariance matrices, a signal fixture, source pins, results, and reproduction instructions. These results concern the measurement method. They do not estimate neural source count or clinical benefit.

SEEG · common-average reference · participation ratio · contact variance · covariance · dimension estimation · reference sensitivity