Invertible contrast changes within shafts preserved the Gaussian score between shafts in positive definite 1–80 Hz SEEG covariance from 36 patients

We tested Gaussian dependence scores in public stereoelectroencephalography (SEEG) data from two interictal records in each of 36 patients. The source was OpenNeuro ds004100, version 1.1.3. In adjacent bipolar contrasts, the mean within-shaft share was 66.7% (95% patient bootstrap interval 63.7% to 69.7%); this covariance share is not an artifact fraction. Three contrast bases retained the same signal subspace. Mean total scores were 0.340, 1.077, and 0.748 nats per channel for adjacent, anchor, and orthonormal contrasts, while each basis gave the same mean between-shaft score of 0.119 nats per channel. A determinant argument proves this equality for invertible changes within fixed shaft blocks and positive definite covariance. We also derive an exact total-correlation score for adjacent differences of independent Gaussian contacts. In the 30–80 Hz band, all 36 observed patient means were below the diagonal contact model means; the model cannot identify an artifact fraction. The scores describe zero-lag covariance and do not show causal communication. This secondary analysis reuses an existing cohort. All 72 compact covariance inputs were rebuilt exactly from public source ranges, and the complete score analysis reproduced seven output files exactly.

stereoelectroencephalography · SEEG · Gaussian total correlation · bipolar montage · covariance · electrode shafts · contrast basis · independent Gaussian contacts