Personal alpha spatial filters had no clear mean balanced eye-state accuracy advantage over peer filters after one month in 60 adults with personal thresholds

We tested personal and peer alpha spatial filters in 60 adults from OpenNeuro ds004148 v1.0.1. Both methods used personal thresholds from the initial session. Weights and thresholds stayed fixed at the one-month test. Mean balanced accuracy was 89.68% for personal filters and 88.73% for peer filters. The paired difference was 0.95 percentage points. The 95% bootstrap interval from 10,000 samples was [-0.25, 2.92] points. All peer fits were repeated for each sample. The interval includes zero and does not prove equivalence. Bootstrap samples used fewer distinct peers and shifted the mean difference upward by 0.33 points. Fixed state order and different released state processing limit the result. The package preserves 360 feature records, source versions and hashes, retrieval code, all refit metrics, figures, and complete repeat instructions.

EEG · alpha rhythm · spatial filters · personalization · balanced accuracy · test-retest · bootstrap