Output phase tests gave excess beta phase slope index flags under independent contact phases in two public SEEG spectral models

We tested phase slope index (PSI) phase tests with fixed spectra from two public SEEG records. The model kept recorded-contact magnitudes and drew independent contact, segment, and frequency phases. The sample had 28 contact quartets in total. Each record had 200 two-second Hann segments. A flag was a test rejection at the 5 percent level. In the primary 13–30 Hz band, phases drawn after bipolar subtraction gave 6.98 percent flags with equal patient weights. Phases drawn before subtraction gave 4.98 percent. Their paired difference was 2.01 percentage points, with a Monte Carlo standard error of 0.22 points. The separate-contact control gave 5.11 percent output flags. We proved zero population PSI and exact finite-sample variance identities. Equal-power checks showed conservative output tests at small segment counts and excess flags at large counts. The package preserves exact inputs, all primary null scores, source locks, code, results, and checks. The result concerns this conditional spectral model. It does not estimate neural or clinical error rates.

phase slope index · surrogate testing · bipolar montage · stereoelectroencephalography · phase randomization · spectral connectivity · Monte Carlo calibration