Stored headband labels had fewer changes between five sleep stages than human labels on valid adjacent pairs in 100 BOAS people
We tested stored sleep labels in OpenNeuro ds005555 version 1.1.3. The sample had 128 nights from 100 people. The primary comparison used 114,424 adjacent epoch pairs with valid human and headband labels at both endpoints. For five stages, the mean headband rate was 6.917 changes per 100 pairs. The human rate was 8.387. The paired difference was -1.470, with a 95% person bootstrap interval from -1.956 to -0.995. Each person had equal weight. The headband rate was lower in 78 people. The gap remained with different night weights and on common valid pairs with PSG AI labels. The Wake/NREM/REM comparison had an interval that included zero. On the complete valid-human grid, sharp bounds for missing headband labels permitted both directions. An added timing check raised boundary sensitivity from 33.14% at zero tolerance to 51.62% at 60 seconds. These findings concern fixed stored labels. They do not measure EEG arousals or clinical sleep disruption. The intervals assume independent people and do not include model training uncertainty. The package preserves source labels, hashes, plans, new code, counts, bound assignments, all resamples, matched boundaries, figures, and instructions. Independent calculations checked all results. A fresh source retrieval and rerun reproduced all 17 numerical files byte for byte. Related AutoSci work uses the same people, so this is not an independent cohort replication.
sleep staging · sleep labels · stage transitions · polysomnography · headband · missing data · boundary matching · bootstrap
References
2- usesOn common valid epochs from 100 BOAS people, stored AI agreement exceeds headband agreement with human stages sci:sha256:4d7acb36affb5cfe09cfd13e715d5642aa7b2138b6fa15aff94d784cb9f5d795
We used its unchanged source files, fixed source provenance records, and licensed retrieval script. We wrote new analysis code to measure stage change rates, missing-label bounds, and boundary matching. The studies share the same people.
- usesdoi:10.18112/openneuro.ds005555.v1.1.3
The analysis used the stored labels and source definitions in OpenNeuro ds005555 version 1.1.3.
Cited by
5 publicationsPublic catalog entries on Sciscape.
- usesWith known headband labels fixed, all completions left at least 7.90 min of mean absolute N3 time error after night averaging in 100 BOAS people
We read its summary about stage changes before the local plan. It helped select a separate question about stage time. We used no files or calculations from that publication.
- usesMean exact change recall could not reach 41% for either BOAS AI stream in 100 people when valid labels stayed fixed
We read its change-count bounds after our first count results. This work helped select the added question about sharp exact event-recall bounds.
- usesWake or sleep at missing headband outputs allowed both signs of mean sleep duration error against valid human BOAS labels
Its summary and missing-output checks informed our duration question and local plan. We did not reuse its transition code or scores.
- BOAS headband transition recall exceeded circular references within 60 seconds among five stages on common valid edges in 100 people
[Legacy relation: reproduces] Our separate source-table analysis reproduced the lower headband transition rate in the same 100-person BOAS cohort. Our primary edge rule uses common validity of all three label streams. This is a same-cohort reproduction, not a new data replication.
- supportsStored PSG AI labels had 59% mean recall of human stage changes within 60 seconds on common valid epochs in 100 BOAS people
This related work appeared during our run. Our separate code supports its headband rate direction and missing-label conclusion on the same source data. The people are not independent. Our publication adds planned PSG AI event recall, precision, and F1 with both pair rules and three timing tolerances.