A common contact caused positive beta power correlation in two SEEG spectrum models with independent contact phases

This study tests raw beta power correlation in two public SEEG spectrum models. It uses 400 recorded contacts from OpenNeuro ds005670, version 1.0.0. Each model fixes the recorded Fourier magnitudes. It gives every contact and frequency bin an independent uniform phase. A common contact gave exact mean correlations of 0.245865 and 0.246217. Channels with no common contact gave zero predicted covariance. There were 199 complete phase draws for each of three windows and four bands in each patient. The draws confirmed the exact predictions. A source quality sensitivity check retained the result. The package gives an exact finite-spectrum covariance formula and a NumPy tool for real montages. It preserves prepared signals, source versions and hashes, code, all control statistics, figures, and checks. Separate source extraction reproduced all saved signals exactly. A full offline repeat gave identical results. The formula gives ensemble correlation. It does not give the exact expectation of finite-record sample Pearson correlation. This evidence covers the stated model and two short resting records. It does not identify neural coupling or a patient population effect.

SEEG · power correlation · Fourier phase randomization · finite-spectrum covariance · bipolar montage · signal processing