Phase randomization order changes 1–40 Hz correlation thresholds in two 180-second SEEG samples

We tested phase randomization before and after bipolar subtraction in public SEEG data. The source was OpenNeuro ds005670, version 1.0.0, from two patients with epilepsy. We used seven 20-second windows from the first 180 seconds of each recording. The tested band was 1–40 Hz. For shared-contact pairs, median 95% absolute-correlation thresholds were 0.607 versus 0.137 in sub-01. They were 0.616 versus 0.134 in sub-02. Threshold crossings changed from 20.4% to 80.4% and from 27.6% to 89.1%, respectively. Disjoint-pair thresholds also changed because the procedures preserve different spectra. Observed shared correlations were mostly positive, while the contact-independence null was negative. A secondary signed test therefore gave different decisions from the upper absolute-correlation test. A simulation with independent recorded contacts checked the null assumptions. The package preserves all primary spectra, code, seeds, results, numerical checks, metadata, and source retrieval recipes with hashes. The contribution is a reproducible method benchmark. It does not establish neural connections or a population effect.

stereoelectroencephalography · SEEG · bipolar montage · phase randomization · correlation · surrogate null models · shared-contact pairs · signal processing