Whole-cycle shifts kept null rejection below 5% in circular phase controls from 35 SEEG records with periodic gaps

We tested a correction for false power dependence after shared voltage gaps. We used spectra from 35 public SEEG records from 21 people. We made independent circular phase controls. Each control had four channels and four seconds. A shared mask removed nominally 50 ms every 500 ms. The score was the mean power correlation over six pairs. At a 5% test level, arbitrary time shifts rejected 61.88% of independent controls. Shifts by whole gap cycles rejected 4.48% (95% subject bootstrap interval 4.04–4.91%). The full corrected test detected 62.86% of artificial controls with a 0.50 shared voltage variance fraction. A proof gives a valid finite-group test under independent circular sources, an exactly periodic mask, and processing that commutes with cycle shifts. A single pair had only eight relative shifts. Its minimum p value was 0.125. The joint test can reject in four seconds, but it cannot identify a pair. The package keeps code, crops, hashes, full-source retrieval instructions, all draw results, and figures. All 35 fresh source crops matched. A second run gave equal numerical arrays. This is a method test on reused source spectra, not new evidence of neural connectivity or live interface performance.

stereoelectroencephalography · signal processing · power correlation · circular shifts · periodic gaps · finite-group test · null calibration