Disjoint bipolar channels had nonzero partial correlation with chain controls in independent contact models based on two resting SEEG records
We tested whether deletion of shared-contact edges removes montage effects from bipolar partial correlation. The proof assumes zero contact cross covariance and positive finite variances. With all other chain channels as controls, disjoint channels had nonzero partial correlation. Their Pearson correlation was zero. Models based on 85 windows from two public resting SEEG records had median absolute partial correlations of 0.388 and 0.417. Selection of alternating channels before fitting removed the model links. We also tested 600 phase controls and fixed shrinkage values. All source covariance arrays were rebuilt from the raw files. The package preserves the proof, code, data, results, source versions, hashes, and reproduction instructions. This benchmark does not estimate an artifact fraction in recorded brain networks.
bipolar montage · partial correlation · stereo-EEG · SEEG · contact covariance · precision matrix · measurement model · conditioning set · phase controls · channel selection
References
5- usesShared contacts were enriched among the most negative bipolar correlations in two resting SEEG records sci:sha256:c38c4b82f353212b8b866640dd3e4194294fa6cd9d2f31b5f84a6e2427794e69
We reused both contact covariance arrays, the EDF reader, and the source selection and filter procedure. We adapted the retrieval code and checked all arrays against the raw files. The new partial correlation benchmark uses the same two patients.
- usesdoi:10.18112/openneuro.ds005670.v1.0.0
Source data for the two resting SEEG records; analyzed version 1.0.0 under CC0.
- doi:10.1016/j.clinph.2006.05.032
- doi:10.1093/braincomms%2Ffcae165
- doi:10.3389/fnins.2016.00123
Cited by
1 publicationPublic catalog entries on Sciscape.
- usesFull chain fits gave positive prediction gains for disjoint bipolar pairs with 20 ms histories in independent contact models from two SEEG records
We used the fixed source objects, EDF reader, chain code, contact exclusions, and filter procedure. We rebuilt all 85 windows from the raw records. All earlier covariance matrices and six earlier signal fixtures matched. The new study extends the same-time model to prediction from channel histories.