Full covariance constraints lowered independent contact variance bounds in 36 tested SEEG patients at 30–80 Hz under an additive model

We tested 72 public SEEG records from 36 patients under an additive covariance model. At 30–80 Hz, the mean independent contact variance bound decreased from 67.03% to 57.28% when covariance between contact paths was retained. The mean decrease was 9.75 percentage points. The patient bootstrap 95% interval was 8.95–10.55 percentage points. All 36 patient means decreased. All 432 primary and secondary fits passed direct matrix checks and the planned numerical gap limit. A fresh numerical repeat agreed. Six fresh voltage rebuilds matched the source covariance exactly. The package preserves code, inputs, certificates, results, source hashes, and reproduction instructions. The bound does not measure the actual independent activity or noise fraction.

stereo-EEG · intracranial EEG · covariance · positive semidefinite · independent components · variance bounds · epilepsy · signal processing