A fixed SEEG reader gave 6.04 percent mean error reduction against a reference with the measured baseline after 20 calibration trials in six people

We tested whether earlier SEEG voltage improves a later pulse waveform forecast. The source was public OpenNeuro ds006065 version 1.0.0. Six people supplied eight blocks and 439 later test trials. Each fixed ridge reader used the first 20 pre-block trials. The input covered 150 ms. The target covered 50 to 250 ms after the inferred pulse. The reference included the measured input baseline. The mean error reduction was 6.04 percent. Its descriptive 95 percent person bootstrap interval was -0.28 to 15.20 percent. The exact sign-flip value was 0.21875. Four of six person means were positive. This test did not establish a consistent gain. Matched windows before pulses gave 7.02 percent mean reduction. Their gain was higher in all six people. This comparison is descriptive. An exact proof shows that the same baseline cancels from target and forecast error. An independent-input simulation showed 45.93 percent apparent gain against a constant corrected waveform. Its gain against the reference with the measured baseline was -1.88 percent. Thus, apparent gain can occur without future information. The package preserves four figures, compact inputs, models, predictions, full score tables, code, and retrieval instructions. All 16 complete source hashes were checked. A separate run reproduced all 12 forecast result files. The analysis uses reused clinical records and inferred timing. It does not test a causal neural response, a live controller, or clinical benefit.

SEEG · ridge regression · waveform forecasting · baseline correction · brain stimulation · electrophysiology · forecast evaluation