A mean 4.56 percent of top 10 percent raw beta PSI edges kept opposite reference signs across two minutes in 19 SEEG records

We tested raw phase slope index (PSI) signs in 19 public SEEG records from OpenNeuro ds003876 version 1.0.2. Each record supplied two adjacent minutes. We selected the top 10 percent of absolute beta PSI values in the first minute under the stored reference. We compared the same edges under the stored reference and common average subtraction. A conflict required the same nonzero sign across minutes in each reference, with opposite signs between references. The mean conflict fraction was 4.56 percent per patient. Its 95 percent patient bootstrap interval was 3.36 to 5.95 percent. All 19 patients had at least one conflict. The interval assumes independent patient sampling and describes this selected, reused cohort. An exact delayed-pair model showed a population sign reversal after common average subtraction. The model included an independent third channel with higher power. It was not fitted to patient data. All source ranges were checked. Rebuilt Fourier coefficients and 23 analysis files matched exactly. The package includes code, inputs, provenance, all scores, figures, and reproduction instructions. Raw ranks and two equal signs do not establish significant neural connections or causal flow. No chance baseline or edge significance test was used.

phase slope index · stereo-electroencephalography · voltage reference · common average · signal reproducibility