A new fit lowered mean 50 ms phase forecast error by 1.10 degrees against a centered theta reference in seven SEEG rest records

A fit from pre-stimulation rest had 58.05 degrees of mean error on later post-stimulation rest. A new fit used early post rest with a 56 second calibration cutoff interval. Its mean error was 56.95 degrees. The mean gain was 1.10 degrees. The 95% descriptive person bootstrap interval was 0.20 to 2.50 degrees. The exact two-sided person sign-flip test gave p=0.06250. Six of seven people had lower error. The descriptive bootstrap interval did not include zero. The sample has only seven people. The exact sign-flip test and the descriptive bootstrap interval use different assumptions. The result does not establish a gain in a new person. The source was OpenNeuro ds006065 v1.0.0, with four fixed SEEG pairs per person. The forecast used only past supplied samples. Its centered theta phase reference used future samples. The package includes extracted voltage, forecasts, model weights, source hashes, figures, and repeat checks. The result does not show that stimulation caused the change or that a live device improved.

SEEG · phase forecasting · theta oscillations · ridge regression · brain-computer interfaces · calibration · OpenNeuro