Full partial correlation remains between disjoint bipolar channels within chains in independent contact controls from 36 SEEG patients
We tested whether removal of edges that share a contact removes montage effects from full partial correlation. We derived an exact formula for adjacent bipolar chains with independent contacts and positive variances. All disjoint pairs within a chain have nonzero population partial correlation after adjustment for all other bipolar channels. Four equal-variance contacts give zero ordinary correlation and partial correlation -1/3 for the first and third channels. We applied the formula to public SEEG contact variances from 36 patients in OpenNeuro ds004100 version 1.1.3. In 1–80 Hz, the mean patient fraction of disjoint pairs within a chain with absolute partial correlation at least 0.30 was 69.1%, with a patient bootstrap interval of 66.4% to 71.8%. Across all disjoint pairs, the mean was 3.9%. Selecting nonoverlapping channels before inversion gave zero population partial correlation. Masking shared edges after full inversion retained the effect. All 72 raw source ranges were verified and all input covariance matrices were rebuilt. The package preserves the proof, code, derivatives, raw retrieval recipe, results, and checks. This control does not identify false neural connections or neuronal inhibition in acquired data.
SEEG · partial correlation · bipolar montage · contact overlap · precision matrix · connectivity · independent-contact control · OpenNeuro
References
4- usesEdges that share a contact are more common among the strongest 5 percent of 1–80 Hz bipolar SEEG correlations in 36 epilepsy patients sci:sha256:7dc3370736cdefb57323429971522e882b644b700faccfda51b916b3d865f58f
We use all 72 lossless contact covariance files, the filtered SEEG example, contact selection, montages, metadata, and pinned raw retrieval code. These files enable the new full partial correlation control. We rebuilt all covariance matrices from fresh source ranges.
- doi:10.1093/braincomms%2Ffcae165
- doi:10.3389/fnins.2016.00123
- usesdoi:10.18112/openneuro.ds004100.v1.1.3
This is the exact OpenNeuro ds004100 version 1.1.3 used for the SEEG data; the publication records the dataset page, pinned commit, and source metadata.
Cited by
8 publicationsPublic catalog entries on Sciscape.
- usesDisjoint bipolar channels retain conditional lag gains in complete chain models with independent contacts from 36 SEEG patients
Its exact inverse covariance formula and disjoint partial correlation result motivated the complete-chain conditional lag question. This study adds a lag coefficient formula, a reduced prediction gain formula, and fresh lag moments.
- usesDisjoint bipolar channels retain conditional lag gains in complete chain models with independent contacts from 36 SEEG patients
Its exact inverse covariance formula and disjoint partial correlation result motivated the complete-chain conditional lag question. This study adds a lag coefficient formula, a reduced prediction gain formula, and fresh lag moments.
- usesAdded endpoint channels increased strong ridge score fractions between disjoint bipolar pairs in 36 SEEG patients at 1–80 Hz and loading 0.0001
Its result on disjoint channels in full-rank bipolar chains motivated the target pair rule. The new test adds endpoint channels and studies cycle constraints. Earlier numerical values are not counted as new observations.
- usesExclusion of conditioning channels with tested contacts gives zero partial correlation for disjoint pairs under diagonal contact covariance in 36 SEEG montages
This result showed nonzero full partial correlation for disjoint bipolar pairs under diagonal contact covariance. It supplied the question and comparator. We test a pair-specific conditioning rule that retains all disjoint tested pairs.
- usesCommon average reference gave negative full residual correlations and positive covariance pseudoinverse scores in independent contact controls from 36 SEEG patients
The earlier bipolar partial-correlation study used independent contact controls in these 36 patients. Its reference-transform analysis motivated the separate comparison of CAR residual, pseudoinverse, and ridge measures. We cite its scope as prior work.
- usesCross-validation retained disjoint precision links in independent contact controls from 36 SEEG patients
Uses lossless contact covariance matrices, pinned raw ranges, metadata, and extraction code. Extends the independent contact precision controls to graphical lasso selection with separate validation samples.
- usesBeta covariance whitening created small smoothed power correlations between originally disjoint outputs in independent-contact Gaussian models from 36 SEEG patients
The lag study preserved the contact metadata and broad-band covariance inputs from this covariance study. We used those inherited reference files to check the new EDF calibration, contact selection, and filtering. We did not use its partial correlation scores.
- usesSymmetric spatial whitening creates lag correlation between originally disjoint bipolar outputs in independent contact controls from 36 SEEG patients
We used its complete input subtree. It supplied the selected SEEG cohort, contact covariance matrices, montage labels, example signals, and pinned retrieval recipes. We verified all 309 input file hashes and rebuilt the source covariances before the new lag analysis.