Common references changed the 30–80 Hz power slope split while bipolar slopes stayed fixed in two 120-second resting SEEG sections

We tested whether a common reference changes a spectral component split. We used seconds 130 to 250 of both resting SEEG records in OpenNeuro ds005670 version 1.0.0. The slope fit used 30 to 80 Hz and removed 48 to 52 Hz. With identical linear processing at both endpoints, bipolar power stayed fixed. Across recorded, contact-average, and shaft-average references, the mean ratio allocation ranged from 81.8 to 92.6 percent in sub-01 and from 92.1 to 98.7 percent in sub-02. The median per-path range in the endpoint exponent change was 0.1022 and 0.1674. A separate pair midpoint control gave a constant power ratio and allocated the full span change to endpoint power. It cannot define one global montage. Both time halves and both alternative fits retained reference dependence. The empirical scope is two short sections from patients with epilepsy. A large ratio allocation alone cannot identify reference-independent neural contributions. The package supplies a proof, phase-preserving Fourier data, source code, pinned full-data retrieval, metadata, results, figures, and repeat instructions. Independent decoding, direct Fourier phase checks, regression checks, and exact source rebuilds passed.

stereo-electroencephalography · SEEG · spectral slope · power spectrum · common reference · bipolar derivation · cross-spectrum · reproducibility