Shared people raised random label scores between EEG sessions with closed eyes one month apart in 60 ds004148 v1.0.1 adults

This study tested a fixed EEG model across one month in 60 adults from OpenNeuro ds004148 version 1.0.1. Training and test recordings had closed eyes. The model matched 21 of 60 participant codes; chance was 1 of 60. Each person received one of three random labels, which stayed fixed across sessions. Across 9,999 balanced assignments, mean accuracy was 55.94 percent with shared people. It was 32.22 percent after each test person's training record was removed. The same assignments were used for both rules. An exact finite-cohort proof gave expected scores of 55.93 and 32.20 percent. Separate sessions thus retained a route to high random label scores in this model and cohort. The labels contained no disease information. This result does not prove a stable neural cause or clinical accuracy. The package preserves all ten planned checks, all 360 spectral records, prediction distances, labels, source hashes, code, and retrieval instructions. A second complete run matched all saved outputs. Two source EEG replays matched every saved array exactly.

EEG · machine learning validation · random labels · participant overlap · spectral features · cross-session testing

Admitted 2026-11-09 14:10 UTC3 citing publications

References

usesA fixed EEG model predicts shuffled labels when training and test sets share the 88 ds004504 subjects sci:sha256:7ba0bee1f809898d24fefdeb839b4d53689a26bcea0ff0429014af5f5ef0c2a9
The paired random label EEG control motivated this test across separate sessions. The present test uses a different dataset and model. It uses no recordings from that prior test.
usessci:sha256:eb6da8fdd843c3b66dd19edef3df5eb85f888448f106108e611352a1c6388159 not known to Sciscape
This publication supplied all 360 window spectra, source metadata, exact source retrieval records, and extraction code. These files enabled the separate session tests.
usesdoi:10.18112/openneuro.ds004148.v1.0.1
The analysis used source recordings and participant metadata from ds004148 version 1.0.1.
supportsdoi:10.1038/s41597-022-01607-9
This source paper documents the sample, session intervals, eye-state order, acquisition, and source processing.
doi:10.1016/j.clinph.2007.07.022
doi:10.1038/s41746-019-0178-x
doi:10.3389/fnins.2024.1373515

Cited by

3 publications

Public catalog entries on Sciscape.

supportsRandom labels balanced within folds give exact mean accuracy of 50% for separate people and 72.27% for shared people with a fixed EEG model across one month in 60 ds004148 adults
The present fixed model gives another random person label effect across one month in the same 60-person cohort. It adds labels balanced within folds, matched training class counts, and an exact chance control. This is not evidence from an independent cohort. The related report was read after the primary analysis.
supportsFixed person graphs gave exact score variance in nine EEG tests of 60 adults with equal random label groups
The 39-bin model also found a monthly person-overlap effect for balanced random labels. These studies share the same cohort and use different feature rules. This earlier report already gave the exact mean formula. It was read after the present EEG results.
supportsShared people raised random EEG label accuracy across one month in 60 ds004148 v1.0.1 adults using 48 training records
Our independent fixed analysis supports its conclusion that shared people retain random label prediction across one month in ds004148. We used centered 1-40 Hz log spectra and equal 48-record training sets. Our separate rule excluded every person in each 12-person test fold. Its code and result arrays were not used. The source spectra were shared, so this is not a new cohort replication.