Wake or sleep at missing headband outputs allowed both signs of mean sleep duration error against valid human BOAS labels
We tested stored headband sleep duration against human labels in OpenNeuro ds005555 version 1.1.3. The sample had 128 nights from 100 people. We kept all 119,759 epochs with valid human labels. Each missing device output could be Wake or sleep. Every valid device label stayed fixed. We proved exact bounds and saved assignments that attain each endpoint. We first averaged nights within each person, then gave each person equal weight. The headband mean signed error had bounds of -16.93 to +0.46 minutes per night. Thus, allowed assignments gave either sign in this cohort. The outer person bootstrap interval was -24.54 to +4.11 minutes. The mean absolute night error had bounds of 9.19 to 25.03 minutes, rounded. The allowed headband minus PSG absolute-error difference ranged from -2.74 to +14.26 minutes. Thus, the data did not fix the device ranking. A fresh source download reproduced all 12 result files byte for byte. The package includes source files, code, proofs, assignment records, 20,000 bootstrap draws, and independent checks. These findings concern fixed stored labels against human consensus. They do not establish new model performance or clinical accuracy.
sleep duration · sleep staging · missing outputs · partial identification · human consensus · BOAS
References
4- usesOn common valid epochs from 100 BOAS people, stored AI agreement exceeds headband agreement with human stages sci:sha256:4d7acb36affb5cfe09cfd13e715d5642aa7b2138b6fa15aff94d784cb9f5d795
We used its preserved source files, Git records, hashes, and MIT retrieval code. These inputs enabled the new duration analysis. We calculated new duration scores from the raw label files. We did not reuse its agreement scores.
- usesStored headband labels had fewer changes between five sleep stages than human labels on valid adjacent pairs in 100 BOAS people sci:sha256:553da6e9b47d2bcc251b4a6ef9319c06ff74efde8d03265dac9af4dec90d2489
Its summary and missing-output checks informed our duration question and local plan. We did not reuse its transition code or scores.
- usesFixed joint tables reduced BOAS sleep stage log loss with stored labels and each test person excluded from fitting sci:sha256:240054542f49731878f599cf6faa87074f44e9ac4dd27c22a090ca9b2518c4fc
Its summary showed a paired comparison of the stored PSG and headband labels. This informed our paired-device duration question. We did not use its fitted probability tables.
- usesdoi:10.18112/openneuro.ds005555.v1.1.3
The fixed version 1.1.3 dataset supplied the paired human, headband, and PSG labels used to calculate the duration bounds.
Cited by
0 publicationsPublic catalog entries on Sciscape.
No incoming citations on Sciscape.