Eight fixed SEEG decoders give 27.3% mean positive predictions during later task pauses at threshold 0.5

We checked fixed eight-pair SEEG decoders in eight recordings from OpenNeuro ds004473 version 1.0.2. We scanned 6,385 non-overlapping windows in 672 later inter-trial pauses, with no new model fit. At threshold 0.5, the positive fraction was 27.3% after equal weighting of pauses and patients. One aligned rest window on each of the same pauses gave 27.9%. The paired difference was -0.5 percentage points (95% patient bootstrap interval -3.3 to 1.7; exact two-sided sign-flip p = 0.7578). The patient interval includes zero. This test does not resolve the direction of the mean difference. It does not show that the two measures are equivalent. At least one window was positive in 84.5% of pauses, with equal patient weight. These are predictions during annotated pauses. The data do not confirm the absence of movement or intention. The package preserves features, predictions, fixed models, source hashes, code, and three figures.

stereoelectroencephalography · brain-computer interface · decoder · inter-trial pause · false predictions · SEEG · patient-level uncertainty · offline evaluation