A fixed 250-ms SEEG power decoder did not detect a cue versus post-keypress accuracy difference under time holdout in eight patients

We tested whether response side decoding changed between early cue and post-keypress signal windows. We used 1,986 matched trials from eight patients in OpenNeuro ds004473, version 1.0.1. Each window had 250 samples, or about 250 milliseconds. A fixed ridge decoder used beta and gamma power from adjacent SEEG contacts. Five consecutive trial blocks separated training and testing. Mean balanced accuracy was 48.16% for cue windows and 50.51% after keypress. The paired difference was 2.36 percentage points. The 95% patient bootstrap interval was -1.80 to 7.04 points. The exact two-sided sign-flip test gave p = 0.44531. The test did not detect a timing difference. It does not establish equal signals. No window or patient test passed the stated correction for multiple tests. The package adds matched features, source range hashes, raw check fixtures, code, and reproduction steps. The result applies to this fixed decoder and cohort. It does not isolate sensory and motor activity.

stereoelectroencephalography · SEEG · response decoding · ridge classifier · time holdout · beta power · gamma power · conditional permutation · negative result