With fixed labels and known masks, a four-cell table reduced stage change audit variance by 15.0% in 50 BOAS v1.1.3 test people

We tested random pair review of stored sleep stage changes in BOAS version 1.1.3. We assumed fixed reference labels and a known common validity mask. We fitted a four-cell table on 50 people and tested it on 50 other people with 64 nights. The table used both automatic change indicators. At a 10% maximum label cap, the table reduced exact audit variance by 15.04% against direct pair review. Standard deviation fell from 0.65370 to 0.60254 changes per eligible edge hour. Both methods selected 2,858 edges and required at most 5,716 endpoint labels. Raw headband assistance increased variance by 29.87%. The table reduced variance in 47 of 50 test people. All planned cohort checks also showed lower variance. The difference estimator was unbiased for any fixed table under uniform random pair selection. Independent calculations, complete small-design enumeration, and fresh source retrieval passed. The fresh analysis produced 19 byte-identical result files. The package contains source labels, code, an audit tool, proof, results, and reproduction instructions. The evidence concerns these fixed labels and review selections. It does not test new scorer error, population uncertainty, or total review cost.

sleep staging · audit sampling · difference estimator · variance reduction · polysomnography · wearable EEG