A target from independent contacts with equal powers increased mean beta model loss for the next minute in 19 public SEEG records
We compared two covariance shrinkage targets in 19 selected interictal SEEG records from OpenNeuro ds003876 version 1.0.2. All adjacent contact pairs gave 1,773 bipolar channels. One target used the identity matrix. The other used the contact subtraction map with independent contacts and equal powers. Both models used the same measured channel powers. Three validation folds within the first minute selected a weight from a fixed five-value grid. The separate next minute supplied beta model loss. The contact-map target increased equal-patient mean loss by 0.533785 nats per channel and vector. The patient bootstrap interval was 0.401448 to 0.700279. All 19 records showed an increase. Reversed minutes, removed scalp name prefixes, and removed clipped records preserved the direction. Disjoint pairs gave exactly equal model scores. Simulations showed benefit when contact signals met the target assumptions. Independent matrix checks and a full replay passed. Public source ranges reproduced all 4,222,800 saved Fourier coefficients exactly. The package preserves code, data, source versions, hashes, results, proofs, and reproduction commands. The result applies to this target, grid, and short reused sample. It does not identify a neural cause or test clinical prediction.
stereo-EEG · covariance shrinkage · bipolar montage · beta activity · model loss · OpenNeuro
References
8- usesConditioning on the middle bipolar channel produced beta partial coherence in independent contact models fitted to 19 public SEEG records sci:sha256:8c91b7c24a750154b8c0f9bb7b450a72255d665ca352648f04e0586d5caa1a7a
This publication supplied the 19 contact Fourier arrays, pinned source metadata, EDF headers, byte recipes, and retrieval code lineage. We rebuilt every used array and tested a new covariance target comparison.
- usesReference choice reduced mean 80 to 150 Hz rank agreement more than time change in 19 public SEEG records sci:sha256:16d6a5e5126b007f6f6be4f042d829df32ccd751f41bcfda4c9b9a3f5a80fce2
This publication supplied the EDF helper lineage and original sample selection rules. The reused helper retains this attribution.
- usesSelected covariance between named shafts reduced mean Gaussian beta loss for the next minute in 19 SEEG records sci:sha256:90aac3b00b3aaccfc26e317167e801e07b5f711f0c08a4977abaedee2b6b37a4
This result informed the prediction-loss question in the same selected cohort. It tested covariance between named shafts with fixed shrinkage. We tested two shrinkage targets for adjacent pairs with fit-only weight selection.
- usesdoi:10.18112/openneuro.ds003876.v1.0.2
This is the exact OpenNeuro ds003876 version 1.0.2 identified as the source of the SEEG records used in the analysis.
- doi:10.1038/s41593-021-00901-w
- doi:10.1101/2021.10.15.464594
- doi:10.1016/j.clinph.2006.05.032
- doi:10.1109/TSP.2015.2443726
Cited by
0 publicationsPublic catalog entries on Sciscape.
No incoming citations on Sciscape.