Gaussian controls gave a 20 percent rejection rate for one second shared bipolar beta power under independent contact rotations in 36 SEEG patients

Gaussian contact controls rejected 20.42% of synthetic null observations in the primary one second beta power test, against a nominal rate of 5%; matching contact rotation controls rejected 4.90%. The benchmark used 216 paths from 72 public SEEG files with 36 patient identifiers. Fourier phase controls rejected 20.90%. The Gaussian excess over rotations was 15.51 percentage points (descriptive 95% patient bootstrap interval 11.76 to 19.46), positive in all 36 patient summaries and 182 of 216 paths. An independent seed repeat gave 20.00% for Gaussian controls and 4.95% for rotations. These findings describe errors under a specified synthetic model with independent contact processes and fixed recorded contact patterns. They do not estimate false neural connections, establish that rotations model brain activity, or support clinical prediction. The analysis uses periodic 60 second extracts on a 64 Hz grid from OpenNeuro ds004100 version 1.1.3; the archived participant table lacks HUP151, and no clinical outcome field was used.

stereoelectroencephalography · SEEG · surrogate data · power correlation · finite-sample calibration · null model · contact rotation · Gaussian controls · Fourier phase randomization · epilepsy dataset