Disjoint bipolar channels retain conditional lag gains in complete chain models with independent contacts from 36 SEEG patients

Removal of shared-contact edges after a complete bipolar chain fit left conditional gains between disjoint channels. We proved an exact formula under contact independence. The gain is positive when the target contacts have unequal lag correlations. We used public SEEG from 36 epilepsy patients in OpenNeuro ds004100, version 1.1.3. The model retained contact variances and circular autocovariances and set cross-contact covariance to zero. At 1 to 80 Hz and a nominal 20 ms lag, the mean patient block-median score was 0.000027573 natural log units. The 95 percent patient bootstrap interval was 0.000018658 to 0.000039988. All 36 patient scores were positive. The mean patient fraction of same-chain disjoint edges with gain at least 0.001 was 3.762 percent. Selection of channels with no contact reuse before the fit gave zero off-diagonal gains under the model. The score uses one past value per channel. It does not measure full-history Granger causality or neural direction. The package contains the proof, source code, data moments, all model scores, source records, and a pinned retrieval recipe. All raw feature files and all numeric result files had identical hashes after complete rebuilds.

bipolar SEEG · conditional prediction · lag gain · contact independence · time series · measurement model