A 95% interval included zero for mean 7–13 Hz EEG change error under a first-visit linear age and sex model in 106 ds005385 adults

We tested if an age relation between people predicted later EEG frequency change within separate people. We used pre-task eyes-closed EEG from OpenNeuro ds005385 version 1.0.3. A fixed split gave 101 reference people and 106 test people. The measure was the largest normalized posterior spectral value from 7 to 13 Hz. A linear first-visit model included age and recorded sex. Its age slope was -0.02483 Hz/year, with a 95% bootstrap interval from -0.04065 to -0.00899. Test recording-year gaps were 4 to 7 years. The observed mean change was -0.07311 Hz. The predicted mean change was -0.12904 Hz. Observed minus predicted change was +0.05593 Hz, with a 95% interval from -0.13772 to +0.25434. The primary interval and all seven planned check intervals included zero. The age rule did not establish lower squared prediction error than predicting no change. These results do not prove equal changes or an absence of aging. The intervals assume independent people and condition on the split, sample, and processing. The package saves input spectra, all model draws, person tables, source hashes, code, figures, and reproduction commands. A second full analysis gave identical files. An independent replay reproduced seven saved arrays from four fresh complete source EEG files exactly.

electroencephalography · alpha frequency · aging · longitudinal prediction · bootstrap · Dortmund Vital Study