Exact variance exceeds the independent prediction value for a fixed five-fold three-neighbor EEG test with separate ds004504 v1.0.9 people and balanced random person labels
Repeated EEG windows can give dependent classifier scores when training and test people are separate. We provide an exact mean and variance tool for binary classifiers with fixed one-neighbor or three-neighbor maps. The maps must be independent of the tested labels. The method includes repeated label units and fixed class counts. The main test uses 1,760 windows from 88 people in OpenNeuro ds004504 version 1.0.9. A fixed five-fold three-neighbor test keeps people separate. Exactly 44 people receive each random binary label. Its exact mean accuracy is 49.2591%. Its exact standard deviation is 2.8830 percentage points. Independent Bernoulli predictions with the same mean give 1.1917 percentage points. The ratio is 2.4192. Full enumeration checks 82 fixed-count cases and 14 independent-label cases. Twelve EEG settings pass separate map, overlap, and vote checks. Two fresh rebuilds match all 18 core numerical files. An earlier person-pair method gives identical fractions for all three binary one-neighbor tests. A fresh source download reproduces one person's features with zero error. The other source audits are inherited. The package preserves the proof, code, features, source hashes, random labels, scores, results, and repeat instructions. The results apply to the stated random-label tests. They do not measure clinical prediction or uncertainty across new people.
exact moments · variance · random labels · cross-validation · nearest neighbors · EEG · fixed-count labels · reproducibility
References
6- usesFixed person graphs gave exact score variance in nine EEG tests of 60 adults with equal random label groups sci:sha256:8ff5778349dbb919b9b432b8104403b39231172818215effc56472bab32f0ed1
This archive supplies unchanged person-pair moment code. It independently checks the exact means and variances for the three balanced binary one-neighbor tests.
- usesA fixed EEG model predicts shuffled labels when training and test sets share the 88 ds004504 subjects sci:sha256:7ba0bee1f809898d24fefdeb839b4d53689a26bcea0ff0429014af5f5ef0c2a9
This work made the inherited EEG features, epoch table, source audit, metadata, and extraction code. They support source inspection and feature reproduction.
- usesEpoch shuffles gave 16.6% false rejections at a 5% threshold in five-fold three-neighbor EEG tests with separate ds004504 v1.0.9 people and balanced random person labels sci:sha256:094ac3179208f711e0d759ff2f64583923f1b587933e55acc72dffb0bb2c436c
This archive supplies the fixed EEG benchmark, features, fold rules, and neighbor maps. They support the exact three-neighbor variance calculation.
- usesdoi:10.18112/openneuro.ds004504.v1.0.9
The fixed EEG features and source audit used in the analysis come from this exact OpenNeuro release.
- doi:10.3390/data8060095
- doi:10.1371/journal.pcbi.1006486
Cited by
2 publicationsPublic catalog entries on Sciscape.
- usesShared people reduced expected ridge error from 1.383 to 0.745 at fixed penalty 0.01 for balanced random labels in 88 ds004504 EEG records
This publication supplied prior method context for sign-product moments, fixed class counts, and score dependence across EEG folds. Its exact moments apply to one-neighbor and three-neighbor accuracy. The new implementation applies to quadratic losses from fixed linear prediction maps.
- usesEEG window voting exceeded feature averaging in mean accuracy for fair independent binary person labels in 60 adults tested one month apart with closed eyes
Its fixed binary vote analysis informed the prior work review and the need to keep repeated label units together. We add a comparison of EEG aggregation rules and exact means for arbitrary integer vote counts under fair independent labels. We did not use its data or code.