On 127 BOAS nights, a 10-minute sleep rule gives lower mean headband onset disagreement for every binary completion

We compared two sleep onset rules on 127 BOAS nights from 100 people. Each night starts with a human wake label. We averaged nights within each person and gave people equal weight. A binary completion assigns each missing label to wake or sleep and keeps observed labels unchanged. Exact bounds show lower mean headband onset disagreement with a 10-minute rule for every completion. The decrease is at least 0.339 minutes in this fixed sample. The planned paired comparison uses 95 nights from 80 people. Its change is -0.531 minutes, with a 95% interval from -5.113 to 5.392 minutes. It does not establish a population direction. The package has source tables, hashes, code, a proof, and all bootstrap draws. Separate checks passed. Fresh source files reproduced all 33 numerical outputs byte for byte. Times are from record start. These results concern stored labels, not true physiological sleep onset.

sleep onset · BOAS · missing labels · headband · polysomnography · sleep staging · partial identification · bootstrap · absolute error