Independent choices of 100 selected EEG seconds gave posterior theta-to-alpha AUC limits of 0.637 and 0.964 for 36 Alzheimer and 29 control ds004504 v1.0.9 records with common average reference

This study tested how sample choice changes a fixed posterior theta-to-alpha EEG score. It used all 36 Alzheimer and 29 control records in OpenNeuro ds004504 v1.0.9. The score used a common average reference and pooled powers from ten accepted ten-second epochs. Each person could have a separate choice from every consecutive accepted-epoch sample. Some samples contained grid gaps. The exact attained AUC limits were 0.636973 and 0.963602. The range width was 0.326628. Of 1,044 person pairs, 341 could have either strict score order. A family with one common fractional start had AUC limits of 0.823755 and 0.904215. With 36 positive and 29 negative labels assigned uniformly at random, selection after labels were known had exact expected maximum AUC 0.716827. Any fixed sample choice independent of those labels had expected AUC 0.500000. A proof shows why the independent limits are attained and gives the random-label expectation. Separate code checked all candidate scores and 9,999 random-label draws. Fresh source downloads reproduced three records. A second complete analysis matched all ten scientific output files byte for byte. These results describe permitted selection effects in this finite cohort. They do not estimate clinical test accuracy or the frequency of such selection. The package preserves code, matrices, every candidate, attained choices, source hashes, metadata, one complete offline source example, results, and reproduction instructions.

electroencephalography · sample selection · area under the curve · theta-to-alpha ratio · Alzheimer disease · random labels · signal processing