A shared phase maximum test gives 4.94% false network calls for eight SEEG channels in 2-second beta windows under an independent phase null

We tested 35 public SEEG records from 21 participants in OpenNeuro ds006519 version 1.0.0. Each record gave a network of eight separate electrode shafts in a 2-second beta window. Independent channel phases kept each recorded spectrum fixed. At a nominal level of 5%, phase pair tests called 76.38% of null networks. The marginal pair error rate was 4.9975%. A shared phase maximum test called 4.94% of networks. Its 95% participant interval was 4.85 to 5.03%. Phase tests with Bonferroni correction called 4.72%. Native sample IID tests called every test network, with or without Bonferroni correction. The result controls a fixed network under the stated global null. It does not prove true edges or BCI performance. The package preserves the plan, source code, spectra, all independent test scores, two figures, and source retrieval instructions. A fresh raw rebuild matched all 35 spectrum files. A full replay reproduced 455 score arrays and eight tables exactly.

SEEG · phase randomization · multiple testing · network calibration · beta band · brain-computer interface