Circular shifts after baseline subtraction gave excess beta test rejections under raw power shifts with trial baselines held fixed in eight SEEG patients

We tested whether the order of baseline subtraction and circular shifts changed a beta power correlation test. We used 1562 trials and 831 adjacent bipolar SEEG pairs from eight patients in OpenNeuro ds004473 version 1.0.1. The model held before power and trial baseline power fixed. It drew raw late power from all circular shifts with equal probability. Shifts after subtraction rejected 87.49% of model observations. Shifts before subtraction rejected 4.23%. The mean paired difference across patients was 83.27 percentage points. One patient had 19 trials and could not reject at the 5% level. A constant baseline gave identical P values for both test orders. An exact covariance proof explains the baseline relation preserved by each order. These are test results under the stated model. They do not estimate a neural false-positive rate. The package preserves code, input hashes, trial decisions, all shift results, raw examples, a full retrieval recipe, and reproduction instructions. All 46 numerical files matched in a repeat run. Independent raw checks passed for 32 windows.

stereoelectroencephalography · SEEG · beta power · baseline subtraction · circular shifts · permutation tests · correlation · signal processing · statistical methodology