A fixed table reduced error in sleep stage share differences between records from 20 BOAS v1.1.3 people on valid label pairs
A probability table was fitted to 80 BOAS people with one record each. It was tested on 48 records from 20 other people in OpenNeuro ds005555 v1.1.3. The test used all 36 record pairs within people. It kept only valid human and headband AI labels on common epochs. Mean squared error in five stage share differences fell from 39.18 to 29.48 squared percentage points. The reduction was 9.70, or 24.8%. Its 95% bootstrap interval was 3.20 to 18.00. The bootstrap resampled training and test people separately and refitted the table in every draw. It assumes independent people and fixed source labels. Seventeen of 20 people had lower error. A proof shows that any row probability table cannot increase total variation between share rows. Lower movement alone does not prove lower reference error. The check with at least 95% label coverage gave a reduction of 3.90, with an interval of -0.31 to 9.99. The files lack recording dates. The study does not establish complete night or later night performance. The package preserves source files, code, all bootstrap draws, independent checks, and an identical repeat from fresh downloads.
sleep staging · stage share calibration · within-person differences · probability table · bootstrap · BOAS · OpenNeuro
References
4- usesThe 95% interval included zero for one-record calibration gain over a population table in 20 BOAS people with a fixed 50-epoch prior sci:sha256:0e0d6473508d6811f00ff796b0bcf67d4d0a16627dae5b3582b553ad2c16b10f
This study supplied the cohort split and the source date limit. No outcome values enter the new calculation.
- usesOn common valid epochs from 100 BOAS people, stored AI agreement exceeds headband agreement with human stages sci:sha256:4d7acb36affb5cfe09cfd13e715d5642aa7b2138b6fa15aff94d784cb9f5d795
This archive supplied the unchanged source labels, metadata, and source manifest.
- usesFixed ridge calibration reduced mean squared error of sleep stage shares compared with raw labels on valid pairs in 100 BOAS adults sci:sha256:52d2e2909dd94b72add575c8e85ca168fe70fdbd9654f0e3184112daf6c51485
This study supplied the equal-person probability table definition and the separate record share outcome.
- usesdoi:10.18112/openneuro.ds005555.v1.1.3
The study uses stored headband labels, human consensus labels, and metadata from OpenNeuro BOAS v1.1.3.
Cited by
2 publicationsPublic catalog entries on Sciscape.
- usesConstrained forward correction increased sleep stage fraction mean squared error on valid pairs in a fixed BOAS v1.1.3 test of 50 people
Its study of sleep stage fractions helped to define the outcome and the valid-pair scope. The present study tests a different correction rule with separate training and test people.
- usesConstrained forward correction increased sleep stage fraction mean squared error on valid pairs in a fixed BOAS v1.1.3 test of 50 people
Its study of sleep stage fractions helped to define the outcome and the valid-pair scope. The present study tests a different correction rule with separate training and test people.