Earlier posterior alpha reduced month estimation error in 60 adults with closed eyes under linear adjustment for nearby bands

We tested whether earlier posterior alpha added information after nearby frequency bands in both records were known. We used 60 adults from OpenNeuro ds004148 version 1.0.1. The primary scores used closed eyes, mean window spectra, a mean reference, and 8-13 Hz. Fixed linear models excluded each test person from fitting. Mean squared error fell from 5.324 to 2.565 squared dB when the model also used earlier alpha. The decrease was 2.759 squared dB, or 51.8%. Its approximate 95% refit participant bootstrap interval was 0.843 to 6.188 squared dB. All six planned checks had positive point gains. The median-spectrum check had an interval that included zero. The method requires nearby bands from the month recording. It estimates alpha after that recording. It does not establish a pure neural oscillation or clinical value. The package preserves 180 spectral records, source hashes and metadata, all model outputs, 10,000 shared participant draws, code, and reproduction instructions. Two fresh source files reproduced all saved arrays exactly. A separate fitting method and a complete second run confirmed the calculations.

EEG · posterior alpha · test-retest · prediction error · spectral power · participant bootstrap