Phase randomization after bipolar subtraction gives excess alpha coherence detections under independent contact models from two SEEG samples

We tested two phase null models with contact spectra from the first 131.072 s of two public SEEG recordings. Test draws had independent phases at each original contact. At a nominal 5% level, randomization after bipolar subtraction rejected 99.84% and 99.85% of edges with a shared contact. Randomization before subtraction rejected 4.95% and 4.74%. Edges without a shared contact stayed near 5%. The results held in three sensitivity checks. The package supplies a calibration benchmark, contact maps, derived data, code, and a source retrieval recipe. These are model results. They do not establish false neural connections in the recordings.

SEEG · intracranial EEG · bipolar montage · coherence · phase randomization · surrogate data · shared-contact dependence · null model calibration · alpha band