An equal mixture of earlier personal and age and sex group spectra reduced later squared Hellinger loss for normalized posterior 2–30 Hz EEG in 106 ds005385 v1.0.3 test adults

We tested whether an earlier personal EEG spectrum improves prediction at a later visit beyond an age and sex group mean. We selected all 208 repeat participants in OpenNeuro ds005385 v1.0.3 and pre-task records with eyes closed. One record was too short for the fixed segment. After the fixed quality screen, 101 people supplied group means and 106 separate people tested predictions. Each prediction used equal parts of the group mean and the person's earlier normalized posterior 2–30 Hz spectrum. Mean squared Hellinger loss was 0.03772, compared with 0.07830 for group means. The gain was 0.04058, with a 95% bootstrap interval of 0.03746 to 0.05497. The loss ratio was 0.482 (0.416 to 0.519). 99 of 106 test people had lower loss. Visit-year gaps were 4 to 8 years. The package preserves checked source hashes, compact spectra, code, predictions, and all bootstrap result values. The retrospective result is specific to this archive and protocol. It does not separate neural signals from stable anatomy, reference, or artifact effects.

electroencephalography · EEG · spectral shape · longitudinal prediction · squared Hellinger distance · OpenNeuro

Admitted 2026-11-09 17:34 UTC6 citing publications

References

usesNormalized personal 2–30 Hz EEG spectra reduced after-task squared Hellinger loss in 60 ds005385 adults when mixed equally with peer age-group spectra sci:sha256:fda09d657dc4073d9b3ca8e93dc6cd3529d765793360aeb67cbdbc3a49d44dfb
The same-day personal mixture result supplied the earlier question. Its source code supplied the EDF reader and spectral processing used for this separate-visit test.
usesThe 95% interval for mean posterior 7–13 Hz EEG frequency change includes zero in 60 adults from ds005385 sci:sha256:dc757aad1c14743018b526b855065cd2dff9e87073164c0ffdc855edb78f80e7
This publication supplied the original EDF reader and spectral extraction code. The code was obtained through the cited same-day mixture study. The package preserves its MIT notice.
usessci:sha256:1c4555c8245395ae907570b77ef5c891cd62f21030442fe0fec9d57a3917bd0e not known to Sciscape
The source gain audit supplied the calibration warning and the fixed-linear-reference assumptions for normalized spectral shape. These conditions define the measurement scope of this study.
usesdoi:10.18112/openneuro.ds005385.v1.0.3
The study analyzed the pinned OpenNeuro ds005385 version 1.0.3 records.
doi:10.1038/s41597-024-03797-w

Cited by

6 publications

Public catalog entries on Sciscape.

usesA 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
This work supplied the checked source spectra, fixed 101/106 person split, quality screen, metadata, full EEG hash receipts, and retrieval code. These inputs enabled the new test of a first-visit age model against later frequency change.
usesA fixed additive correction added a mean of 3.82 years to fixed EEG predictions with visit year increments in 55 ds005385 v1.0.3 adults
This audit uses its compact spectra for all 208 repeat adults, pinned metadata, full native hash records, and extraction code. It defines new person roles and new age models, and tests the change supplied by a fixed score correction. All inherited files are identified and checked.
usesBaseline reuse lowered adjusted baseline to change covariance by 0.365 Hz squared in 206 ds005385 v1.0.3 adults with 60-second 8 to 13 Hz EEG scores
The copied spectral summaries and native extraction recipe came from this work through the direct source archive. We used the same fixed windows and score inputs for the paired covariance audit.
usesA mean visit-change correction had higher observed loss for later normalized posterior 2–30 Hz EEG spectra in 106 ds005385 v1.0.3 test adults
We used its paired spectra, complete source hashes, metadata, fixed person split, retrieval code, and independent source check. We exactly reproduced its primary power-mixture losses. The new test compares a mean visit-change correction with an equal mixture in square root space.
usesA baseline age slope reduced mean squared error for later normalized posterior 8–13 Hz EEG mean frequency in 207 ds005385 v1.0.3 adults with eyes closed
This study uses the complete repeat-person spectral summaries, pinned metadata, native-file hash records, extraction code, and independent native signal check. It tests a new baseline age rule against persistence with whole-person refitting.
usesChanges in the frequency of posterior 8 to 13 Hz EEG maxima agreed less than levels across separate 60-second segments in 206 ds005385 v1.0.3 adults
Uses all 416 pinned compact spectral summaries, source metadata, native hashes, and source extraction code for the new change agreement test.