Independent ±3 dB epoch power gains reversed the Fz theta-to-alpha CLN3–HV contrast using 95 ds008896 records’ first 20 epochs

We tested how epoch power scale affects a pooled EEG theta-to-alpha group contrast. The input was ds008896 version 1.0.1, with 32 CLN3 records and 63 healthy volunteer (HV) records. The primary score used the first 20 retained four-second epochs, an Fz reference, and mean posterior channel power. Theta was [4,8) Hz. Alpha was [8,13) Hz. The observed CLN3 minus HV contrast was +0.167018 log10 units. We allowed independent epoch power gains from 10^(-3/10) to 10^(3/10). Each gain multiplied both bands and preserved the epoch ratio. The sharp contrast limits were -0.115497 and +0.441498. Gain witnesses attained both limits. Zero was first allowed at a symmetric limit of 1.77046191 dB. The mean epoch log-ratio contrast was +0.100355 and remained unchanged. All six fixed analysis rules permitted both signs at 3 dB. We supply a proof, a reusable calculator, verified spectra, gain witnesses, bootstrap samples, and a locked source retrieval recipe. Fresh source files rebuilt all archived arrays exactly. Independent solvers and exhaustive small cases checked the limits. A clean run matched every numerical output hash. These are exact sample limits under the stated gain model. They do not show that these gains occurred in the recordings. The parent baseline results were known when the test was chosen. No age or clinical adjustment was possible.

EEG · theta-to-alpha ratio · sensitivity analysis · epoch power · CLN3 disease · healthy volunteers · linear fractional programming