Missing labels could not raise mean headband change recall above 54.76% within 60 seconds for five stages in 100 BOAS people when valid labels stayed fixed

We tested stored sleep labels from 128 nights in 100 BOAS people. We kept valid model labels fixed and let each missing label take any of five stages. The maximum mean headband recall was 54.76% when event times could differ by 60 seconds. Each event could have at most one match. Each person had equal weight. A complete stage path and matched pairs attained this maximum. An independent graph solver checked every source segment and setting. The package preserves the proof, code, source labels, complete paths, results, and retrieval hashes. The bound applies to fixed outputs in this sample and uses human consensus as its reference.

sleep staging · stage-change recall · missing labels · event matching · partial label completion · maximum recall · OpenNeuro · BOAS · sleep monitoring · human consensus