A fixed ridge model did not detect previous response side from SEEG power before fixation in eight patients under conditional time-block tests

We tested whether SEEG power before fixation contains information about the previous response side. We used 1,954 trial transitions from eight epilepsy patients in public OpenNeuro ds004473 version 1.0.1. Features were bipolar power at 13 to 30 Hz and 70 to 110 Hz. The baseline started a median of 4.55 seconds after the previous response. A fixed ridge model used five time blocks with a 30-second exclusion from training data. Mean patient balanced accuracy was 49.98%. The 95% participant bootstrap interval was 47.31% to 52.52%. A test with 4,999 shuffles conditioned on time block, instruction, and current target side gave p = 0.333. No patient passed Holm correction across eight patients. A circular label test gave p = 0.3396. The response-window control detected side information in one patient after correction. This result limits the stated model and window. It does not show that all history effects are absent. The package preserves the transition table, features, code, source versions, fitted models, folds, predictions, shuffled labels, and null scores. All used raw ranges were retrieved again and matched their hashes. All rebuilt feature files matched the inputs. A second analysis reproduced all 339 result files exactly.

stereoelectroencephalography · SEEG · response history · ridge regression · beta power · gamma power · time-block validation