With fixed forecasts, limits from three recent scores covered 13.5 percent of five-pulse SEEG runs after contact order reversal in eight sessions

We tested error limits for fixed SEEG response forecasts in eight sessions from OpenNeuro ds005448, version 1.0.0. Each of 96 conditions had five initial pulses and five later pulses at reversed recorded contact order. The first two pulses fit the waveform forecast and scale. The next three scores set a fixed maximum. A second rule used the maximum of the three latest measured scores before each later pulse. At 20 to 500 ms, the recent-score rule covered 343 of 480 individual pulses (71.5 percent); the fixed rule covered 326 (67.9 percent). The recent rule covered all five later pulses in 13 of 96 conditions (13.5 percent); the fixed rule covered 31 (32.3 percent). The paired complete-coverage change was -18.75 percentage points, with a descriptive patient interval from -27.08 to -10.42. The individual-pulse change after subtraction of the before-pulse control was +3.13 points, with an interval from -0.42 to +6.46. This does not establish an adjusted population gain. We prove that, with a fixed score function, every complete sequence that passes the recent maximum also passes the initial maximum. Under distinct exchangeable scores after a fixed fit, exact complete-five probabilities are 14.67 and 37.50 percent. The chronological data do not establish that assumption. The package includes source products, pinned retrieval recipes, code, all decisions, three figures, and independent checks of all 960 raw ranges. This is an offline test within conditions. It does not test a live brain interface.

stereoelectroencephalography · electrical brain stimulation · forecast error · sequence coverage · conformal prediction · error limits · epilepsy