Division by 2–30 Hz power changed mean posterior theta–alpha log power correlation from positive to negative in repeated closed-eye EEG from 60 adults

We tested the relation between posterior theta and alpha power in repeated public EEG from 60 adults. We used closed eyes, a mean reference from 59 channels, and four-second windows. Mean correlation was +0.307 for absolute log power. It was −0.617 after division by 2–30 Hz power. The paired change was −0.924. Its 95% person bootstrap interval was −0.978 to −0.869. All 180 records showed a decrease. The absolute positive relation became negative in 159 records. An exact model selected theta, alpha, and remaining power independently from each record. This model gave a mean normalized log correlation of −0.563. It shows that the arithmetic alone can give a negative relation with these power distributions. It does not measure a neural cause or a true coupling value. Each of seven planned checks showed a mean decrease. The general issue with proportions is established. This work adds a public EEG test and an exact finite-model audit tool. The package preserves all 360 input spectra, source hashes, code, results, bootstrap draws, proof, and retrieval instructions. Separate calculations and two complete runs agreed. One fresh source record rebuilt exactly. The result applies to this cohort and the stated method.

electroencephalography · EEG · theta power · alpha power · relative power · normalization · correlation · compositional data · independent product model