Independent bin controls raised scaled maximum test rejection to 12.86 percent in local beta phase models with independent contacts and magnitudes from 16 SEEG recordings

Controls with independent bin phases can reject graphs with independent contact phases when frequency bins have dependence. We tested fixed Fourier magnitudes from 16 public SEEG recordings in OpenNeuro ds004100, version 1.1.3. The beta score averaged pairwise phase consistency across 35 bins and 30 windows. A local Gaussian phase convolution kept contact vectors independent and kept each single-bin phase law unchanged. The scaled maximum test rejected 2,055 of 15,984 targets (12.86 percent) with independent bin controls. Matched local controls rejected 809 targets (5.06 percent) on the same graphs. The paired difference was 7.80 percentage points. Its 95 percent bootstrap simulation range was 7.00 to 8.91 points. This range keeps records and pilot scales fixed. An exact block example proves a score variance change. A matched rank proof gives 5 percent global graph error under the chosen model. Fresh source checks of all 16 inputs and a full numerical repeat passed. The package preserves code, inputs, provenance, results, proofs, and repeat instructions. The local model is a stress test. It is not the exact conditional phase law of tapered SEEG given its magnitudes. This work does not estimate false connections in the original signals.

SEEG · phase consistency · frequency dependence · surrogate tests · graph maximum test · calibration · Monte Carlo · signal processing