In 60 ds004148 v1.0.1 adults with closed eyes, a one-month reference reduced the ratio of continuous to dispersed posterior 8-13 Hz mean squared error with 60 planned seconds and equal accepted counts

We tested whether a separate target changes the error benefit of dispersed EEG samples. The primary test used session 1 with closed eyes in 60 ds004148 v1.0.1 adults. The feature was mean posterior 8-13 Hz window power in dB. The separate target was each person's record one month later. Continuous samples had 60 planned seconds and at least 48 accepted seconds. A uniform dispersed control had the same accepted window count. We calculated its expected mean squared error exactly. The group geometric error ratio fell from 2.466 to 1.421. The paired ratio change was 0.576, with a 95% person bootstrap interval of 0.498 to 0.663. The dispersed control still had lower error. Its mean absolute saving was 2.391 dB squared, with an interval of 1.076 to 4.182. The absolute saving increased by 0.413 dB squared despite the smaller relative ratio. An exact formula shows how the target shift adds error to both rules. All eight planned checks had smaller relative ratios with the separate target. Separate loops checked every numerical result. Two raw records rebuilt exactly. All eight repeated output files matched byte for byte. The package preserves spectra, source hashes, code, results, proof, figures, and instructions. The target is an observed record, not a known true trait. Dispersed selection spans the full 280 seconds and does not imply a 60-second recording visit.

electroencephalography · EEG sampling · alpha band · test-retest reliability · mean squared error · paired bootstrap · posterior power