A fixed beta decoder did not detect previous key side at group level before fixation in eight SEEG records

We tested whether beta power before fixation identifies the previous key side in public SEEG data. The test used 1,959 trials from eight epilepsy patients in OpenNeuro ds004473 version 1.0.1. The window ran from 1.25 to 0.25 seconds before current fixation. A fixed ridge decoder gave 51.02 percent mean balanced accuracy for previous side. The conditional group null mean was 49.21 percent. The adjusted p value was 0.21. Previous accuracy exceeded current accuracy by 2.57 percentage points, with a bootstrap interval from -0.14 to 5.41 points. That paired test also had an adjusted p value of 0.21. Neither planned group test met the 0.05 criterion under its stated null assumptions. The result does not establish absence or equivalence. We reconstructed all source features from pinned public signal ranges and checked sampled epochs with an independent reader. The package preserves code, source identifiers, trial links, models, predictions, null labels, results, and reproduction instructions. It provides a bounded result and a response-history benchmark for these records.

stereoelectroencephalography · SEEG · beta power · response history · key side · ridge decoder · balanced accuracy · temporal validation · epilepsy · conditional permutation