One score per site predicted omitted PAC differences near a 1:4 ratio in 56 SEEG patients and 18 MEG participants

We tested whether one score per site predicts omitted phase-amplitude coupling differences. We used public SEEG and MEG matrices from separate cohorts in one study. Two patients and one MEG participant supplied development. A local plan fixed the test in 56 other patients and 18 other MEG participants. The low frequencies were 7 to 14 Hz. The high frequencies were about four times higher. We subtracted supplied surrogate estimates and fit site scores with five edge folds. Mean error reduction against zero was +29.12% (+24.14% to +34.34%) in SEEG and +7.74% (+3.04% to +14.36%) in MEG. Mean added reduction beyond local PAC, symmetric site strength and surrogate site scores was +17.32% (+13.83% to +21.22%) and +3.40% (+1.73% to +5.60%). Each gain is a fraction of the zero-prediction squared error. Intervals resample people. The shared rule passed, with conjunction p=0.010879. Separate people confirm a new endpoint in one reused study; this is not a new-study replication. These tests concern measured coupling contrasts. They do not identify causal flow, common anatomical order, or stable cycles. Raw recordings are unavailable. The package gives compact inputs, source hashes, all outcomes, code and repeat steps.

SEEG · MEG · phase-amplitude coupling · graph ranking · cross-validation · surrogate data · human neuroscience