Month calibration gain was uncertain for linear prediction of posterior 8–13 Hz EEG eye-state contrast in 60 ds004148 adults
We tested EEG calibration for a month target in 60 adults from OpenNeuro ds004148 version 1.0.1. We predicted posterior 8–13 Hz eye-state contrast with a fixed linear model. Each fit excluded the test person. Short calibration used 90-minute targets from other people. Month calibration used their month targets. The observed mean squared errors were 10.422 and 9.316 dB², respectively. The gain was 1.106 dB². Its approximate 95% interval was −1.205 to 3.752 dB². The interval used 20,000 whole-person draws with complete model refits. All five planned sensitivity intervals also included zero. The gain remains uncertain, and the result does not establish equal error. The package preserves spectra, code, scores, predictions, bootstrap draws, source hashes, and retrieval instructions. Independent numerical checks, two fresh raw-record replays, and a byte-identical complete repeat passed. The scope is this cohort and protocol. The interval assumes independent person trajectories and has no tested coverage.
EEG · calibration · posterior alpha · eye-state contrast · prediction · test-retest · bootstrap · OpenNeuro
References
3- usessci:sha256:eb6da8fdd843c3b66dd19edef3df5eb85f888448f106108e611352a1c6388159 not known to Sciscape
This study uses all 360 unchanged spectral files, source metadata, source hashes, extraction code, and record replay code. These files enabled the new identity-excluded calibration comparison and fresh raw-record checks.
- usesdoi:10.18112/openneuro.ds004148.v1.0.1
Source EEG and participant metadata for the 60 adults; the cached metadata pins the source to version 1.0.1 and states CC0.
- supportsdoi:10.1038/s41597-022-01607-9
Provides the source protocol facts used here: session timing, eye-state order, and prior eye-movement component removal.
Cited by
2 publicationsPublic catalog entries on Sciscape.
- usesLinear month forecasts had higher mean squared error than group means for early-to-late alpha contrast change in 60 ds004148 adults
This study supplied the person-excluded linear forecast and whole-person bootstrap refit design for full-record contrast. We apply that design to early-to-late change. Our full-record context row reproduces its month-linear and training-mean point errors.
- usesThe month alpha contrast error gain interval included zero for offline split EEG in 59 ds004148 adults at 120 seconds per eye state
Its month contrast calibration study informed the fixed offset check and identity exclusion in refitted bootstrap draws. We use separate mean offsets for the latest and split rules, rather than its linear model comparison.