Selection from a reused two-second baseline gave a 2.11 dB lower mean alpha score change than earlier selection in 19 visual SEEG patients

We tested how baseline reuse changes a selected SEEG alpha measure. We used OpenNeuro ds005169 and ds006519, both version 1.0.0. Within each shaft, we selected the greatest fixed alpha score from either the change baseline or an earlier period. The main periods were two seconds long. Reused selection gave a mean score change 2.11 dB lower in 19 visual stimulation patients across 89 recordings. The 95 percent patient bootstrap interval was minus 2.47 to minus 1.71 dB. The mean difference was minus 2.01 dB in 17 motor stimulation patients across 30 recordings. An exact decomposition showed baseline selection costs of 2.22 and 2.08 dB. The corresponding post selection terms were 0.11 and 0.07 dB. A one-second control also gave negative differences when all periods came before stimulation. Separate periods do not prove statistical independence. The result measures a procedure difference, not a causal neural effect. We preserve all input spectra, channel choices, code, source pins, result tables, and reproduction instructions. All 119 retained recordings rebuilt exactly from public signal ranges. A full offline repeat gave identical numerical file hashes.

SEEG · intracranial EEG · baseline reuse · channel selection · alpha score · selection bias · spectral analysis · exact decomposition · visual stimulation · negative motor response