Mean 70–150 Hz power rank agreement was lower between two references than between adjacent ten-second periods in 12 SEEG cases

We compared a common average reference and a local neighbor reference in OpenNeuro ds004100 version 1.1.3. The fixed case set contained the first 12 eligible participant IDs in sorted order. The subset was fixed before signal analysis because of transfer time. We used the same 607 center contacts in both references and two recordings per case. Each recording supplied two adjacent ten-second periods. The power band was 70 through 150 Hz, with 118 through 122 Hz excluded. In run 1, mean temporal rank correlation was 0.969 and mean correlation across references was 0.717. The paired mean difference was 0.252, with a 95 percent participant bootstrap interval of 0.135 to 0.390. The two-sided sign-flip probability was 0.000520. All 12 cases had positive differences in both runs. Run 2 gave a mean difference of 0.217, with an interval of 0.144 to 0.308. This evidence applies to the selected cases and short periods. It does not establish a best reference or clinical benefit. The package contains code, source metadata, source version and byte checks, derived tables, figures, plan amendments, and reproduction commands. Independent EDF and statistical checks passed. A fresh source download reproduced one contact table and block archive with identical hashes.

stereoelectroencephalography · SEEG · reference choice · common average reference · local reference · high-frequency power · rank agreement · epilepsy