Power averaging order gave positive and negative 30–80 Hz exponent changes under common average reference in paired SEEG records from 36 patients

We tested whether the order of power averaging and slope fitting changes a fixed SEEG spectral measure. The source was OpenNeuro ds004100 version 1.1.3. We used two 60 s interictal records from each of 36 patients and 4260 matched contacts. The main fit used common average reference, 12 disjoint 5 s epochs, and 30–80 Hz bins with 55–65 Hz removed. The exponent was minus the ordinary least squares slope of log power on log frequency. The difference was the exponent from arithmetic mean power minus the mean epoch exponent. The patient score was the mean of two record medians of contact differences. Its equal-patient mean was -0.012. The 95% patient bootstrap interval was -0.022 to -0.004. There were 21 negative and 15 positive patient scores. The median patient score was −0.002. The equal-patient mean of the median absolute contact differences was 0.038. The mean signed score was −0.134 under the supplied reference and −0.009 with 10 s epochs under common average reference. An exact log-power gap identity matched all 8520 main contact records within 9.4e-15. Its assumptions are positive power, common bins, equal weights, and a linear fit with an intercept. The difference can have either sign. It does not identify a neural cause or the better estimator. The package preserves epoch power, code, source metadata, raw retrieval recipes, results, and checks. The cohort is reused from cited work. The result has no clinical validation.

stereoelectroencephalography · power spectrum · spectral exponent · Welch method · averaging order · common average reference · epoch analysis