A fixed decoder did not detect information about the previous key in SEEG power before fixation in eight patients

We tested information about the previous key in public SEEG power from eight patients. We used 1,956 trials from OpenNeuro ds004473 version 1.0.1. The power windows ended before fixation and started at least 2.26 seconds after the previous response. A fixed ridge decoder gave 50.2 percent mean balanced accuracy for the previous key. The 95 percent patient bootstrap interval was 47.5 to 52.6 percent. The conditional random-label test did not detect information about that key (Holm p = 0.5504). The comparison with the upcoming key also did not pass the corrected test. These results apply to this decoder, power measure, delay, and patient sample. They do not prove absence of information. The package contains code, feature data, predictions, source locks, raw check ranges, and repeat instructions.

stereoelectroencephalography · SEEG · response history · neural power · ridge decoding · balanced accuracy · baseline activity