BOAS headband labels had higher mean destination diversity across Wake, N2, and REM than human labels in 91 eligible people

We tested stored sleep stage labels in OpenNeuro BOAS ds005555 version 1.1.3. We separated change rate from departure destination diversity. Diversity D is the chance that two distinct departures have different destinations. We averaged D across Wake, N2, and REM, then across people. The fixed count rule retained 91 of 100 people. Mean D was 0.5038 for headband AI and 0.4752 for human consensus. The paired difference was 0.0286, with a 95% person bootstrap interval of 0.0020 to 0.0551. Headband labels had 1.69 fewer changes per retained edge-hour. Stricter count checks retained fewer people, and their diversity intervals included zero. These results describe stored labels. They do not establish brain entropy or performance on new people. The package preserves 773 source files, code, random draws, checks, and figures.

sleep staging · sleep stage transitions · destination diversity · Gini-Simpson index · headband EEG · polysomnography · human consensus · stored labels · bootstrap