Independent contact models with positive gamma variances gave nonzero partial correlations between disjoint channels in complete bipolar chains from 33 SEEG patients

We tested whether two bipolar SEEG channels can have a partial correlation when they have no common contact and the source contacts are independent. We used public OpenNeuro ds004100 version 1.1.3. The test used positive contact variances from 60 seconds in each of two files. It included 33 patients and 392 complete chains per file. The independent contact model gave nonzero partial correlations for every disjoint target pair. With equal chain and patient weights, the mean fraction with absolute partial correlation at least 0.2 was 83.0% in A and 82.1% in B. Two alternating channel groups gave zero model partial correlations within each group. The proof uses standard matrix inverse algebra. The package adds a public SEEG audit and reusable code. In measured gamma data, full chain conditioning gave larger absolute correlations than conditioning within the groups, using the same targets. These changes do not identify their neural or technical cause. We retrieved all 72 source prefixes again. The rebuilt covariance passed the saved tolerances. All threshold fractions were unchanged. A separate archived-input replay reproduced the numeric outputs exactly. The package preserves a proof, three figures, compact inputs, all scores, source metadata, hashes, code, and reproduction instructions. The result does not establish neural connections, decoder benefit, or stimulation targets.

stereoelectroencephalography · bipolar channels · partial correlation · independent contact model · signal processing · conditioning set · brain-computer interface