Same-day calibration gave 88.4% mean month containment for nominal 90% frequency intervals from 280-second EEG in 60 adults

We tested whether same-day repeat changes can set intervals for EEG recorded one month later. We used all 60 adults in OpenNeuro ds004148 version 1.0.1. The primary score was the 8 to 13 Hz posterior band maximum from 280 planned seconds of closed-eye EEG on a 0.25 Hz grid. The interval target was 90%. Each calibration group had 30 people. We excluded each test person and calculated the exact mean over all possible calibration groups. Same-day calibration gave 94.24% same-day containment and 88.37% month containment. Month calibration gave 94.01% month containment. Mean width was 1.039 Hz and 1.394 Hz, respectively. Local interpolation gave 79.12% month containment with same-day calibration. The 60-second check gave 90.43%, and the 95% target check gave 96.57%. The package contains a checked subset-count tool, all 180 input spectra, code, source hashes, tables, saved allocations, and retrieval recipes. All three planned raw EEG replays matched every saved array exactly. Prior studies of these people informed the question. These are fixed-cohort allocation results. They do not establish population or personal coverage.

EEG · conformal prediction · calibration · test-retest reliability · alpha frequency · prediction intervals · OpenNeuro

Admitted 2026-11-09 18:17 UTC2 citing publications

References

usesA same-day alpha interval contained 25 of 30 month repeats in one split of ds004148 sci:sha256:8b369abcee780c1d543dfaf057a76d0b5c37a71d62a1b63db6f7deef537214cd
This prior alpha-power study tested same-day calibration for month prediction in one split of the same people. It informed the calibration question. We use frequency scores and exact averages over every uniform calibration group.
uses95% window bootstrap bounds from 280-second closed-eye EEG contained 40 of 60 month repeats of the 8-13 Hz maximum in ds004148 sci:sha256:76db7bd2a462ca799e13436abd7b765bb1d4c9246c1367b01666c9a15cbc7ff6
This prior frequency study tested bounds made from windows within records in the same cohort. Its reported repeat containment informed the question and the choice of a fixed posterior band maximum. We test calibration from other people and all calibration groups.
usessci:sha256:eb6da8fdd843c3b66dd19edef3df5eb85f888448f106108e611352a1c6388159 not known to Sciscape
This source supplied all 180 closed-eye spectra, quality masks, raw source hashes, metadata, and the unchanged extractor. They enabled the repeat frequency scores and three fresh raw replay checks.
usesdoi:10.18112/openneuro.ds004148.v1.0.1
The analyzed closed-eye EEG records come from OpenNeuro ds004148 version 1.0.1.
supportsdoi:10.1038/s41597-022-01607-9
Source study for the cohort and recording schedule described in the publication.
doi:10.1198/016214506000001437

Cited by

2 publications

Public catalog entries on Sciscape.

usesScaled 8–13 Hz EEG frequency ranges set in ds004148 at a 90% target contained 36 of 39 ds007176 V1 repeats
We used its frequency calibration result for the background discussion. It reports allocation averages within ds004148. This context was added after the new analysis.
usesSpectral sets used fewer candidates but had lower month containment in 280-second ds004148 EEG
This work supplied the prior distance interval comparison and the exact calibration-group counting approach. We reproduce its 94.01% primary month-calibrated containment and extend the calculation to spectral grid sets and random priorities at tied boundaries.