Nested pair selection did not show lower absolute score error at 83 SEEG sites in eight recordings

We tested whether electrode-pair selection inside each validation fold improved an early estimate of later response shape. We used public SEEG stimulation data from OpenNeuro ds004696, version 1.0.1. The test kept 83 sites in eight recordings. It used five early pulses and a 20 to 500 ms window. Each fold used the same pair and mean waveform for early and later scores. Absolute score error was 0.09442 with nested selection and 0.09044 with selection from all five pulses. The all-five-minus-nested contrast was -0.00398 correlation units. Its descriptive 95% person interval was -0.02936 to 0.02154. The primary test did not show lower error. Nested selection reduced the signed score gap by 0.05010. Random partitions supported a selection explanation for part of that gap. This is exploratory reuse of one clinical cohort. It does not test brain-computer interface decoding or stimulation benefit. The contribution retains exact compact inputs, source hashes, historical check receipts, full results, and three figures. It omits GPL-licensed retrieval and analysis code, so the exact numerical replay cannot be run from this SCI package. The archived record reports that a fresh source check and a clean numerical replay passed.

stereoelectroencephalography · electrical stimulation · nested validation · cross-validation · correlation · signal shape