A corrected mean increased average normalized 100–500 ms forecast error in five SEEG recordings with four calibration pulses per site

This study tested mean voltage in public stereoelectroencephalography (SEEG) data from five people with epilepsy. It used 138 stimulation sites and 15 recording pairs outside the stimulation shafts. Four early pulses fit each site. Four later pulses tested it. All forecasts used the same waveform shape. The corrected mean increased average error by 0.76% of calibration signal energy. The descriptive 95% interval for error decrease was -3.56% to 2.51%. Only one person had a decrease. The corrected mean had less error than the plain mean in all five people. Its post-pulse gain also exceeded its matched pre-pulse control in all five people. Neither comparison established a gain over the zero-offset forecast. The package adds voltage epochs from -900 to 500 ms, an exact error identity, and independent checks. It preserves compact data, source hashes, code, model values, three figures, and repeat instructions. These are offline results from reused recordings. They do not establish a neural cause or better BCI operation.

stereoelectroencephalography · SEEG · brain-computer interfaces · voltage forecasting · mean voltage · electrical stimulation · forecast error