Under independent, symmetric patient differences, a test detected no beta AUC gain over baseline across conditions in eight SEEG patients

We tested response-key decoding with beta power in public OpenNeuro ds004473, version 1.0.2. The test used eight SEEG patients, 1,988 response trials, and 822 bipolar pairs. A fixed linear ridge decoder trained in one task condition and tested in the other condition. Each patient score averaged both transfer paths. Mean response-change AUC was 0.50372. Mean pre-cue baseline AUC was 0.48657. The mean paired gain was 0.01715. The exact two-sided patient sign-flip p value was 0.609375. The test assumes independent patients and symmetric differences under the null hypothesis. It did not detect a gain at the planned 0.05 level. A descriptive patient-resampling interval was -0.02385 to 0.06568 AUC units. The result applies to the fixed model, band, windows, and records. It does not prove equivalence or absent beta information. The package contains checked response targets, source powers, fitted models, test scores, resampling assignments, source versions, and reproduction code. An exact complete repeat and independent fit checks passed. A raw source check covered 3,614 powers from one trial per patient. Full raw file hashes were not verified. Cue and key direction are closely related. The test does not isolate a motor cause or predict a future movement.

stereoelectroencephalography · beta power · response-key decoding · condition transfer · ridge regression · area under the curve · OpenNeuro ds004473