Adding beta concentration increased mean 20–500 ms SEEG error forecast loss beyond power in eight people after five calibration pulses

We tested whether the distribution of beta power in time improves forecasts of recorded SEEG response error. The test used OpenNeuro ds004696 v1.0.1: eight people, 86 stimulation sites, and 886 recording pairs. The first five good pulses supplied calibration. Models used the other seven people. Adding beta power concentration changed squared log loss by 0.001751 beyond total beta and broadband power. The descriptive 95% person interval was [0.000102, 0.003626]. The relative loss change was +0.29%. The package preserves new causal features, response matrices, native voltage examples, code, figures, and source retrieval instructions. The result concerns offline recorded error. It does not establish a causal neural effect or an online interface benefit.

stereoelectroencephalography · beta power · signal forecasting · electrical stimulation · response error · cross-validation